docs(P03): complete language-derived extension — v0.4

---ci--- project: atelier phase: 3 milestone: v0.4 status: complete phase_role: execution phase_tag: v0.3.3 requirements: covered: [ATELIER-102, ATELIER-103, ATELIER-104, ATELIER-105] partial: [] ---/ci---
docs(ship): P2 complete — v0.3.2 tagged, release 483
2026-08-05 16:07:21 +00:00 · 2026-08-05 16:01:08 +00:00 · 2026-08-05 16:00:31 +00:00 · 2026-08-05 15:53:16 +00:00 · 2026-08-05 15:51:03 +00:00 · 2026-08-05 14:03:35 +00:00
32 changed files with 4976 additions and 9 deletions
@@ -1,13 +1,15 @@
 {
-  "phase": 0,
-  "stage": "plan",
+  "phase": 2,
+  "stage": "complete",
  "milestone": "v0.4",
-  "phase_role": "pre_execution",
+  "phase_role": "execution",
  "project": "atelier",
  "attempts": 0,
-  "updated_at": "2026-08-05T05:20:00Z",
+  "updated_at": "2026-08-05T06:30:00Z",
  "milestone_complete": false,
  "milestone_branch": "milestone/v0.4-edge-quantum-langs",
-  "phase_branch": "phase/00-pre-execution",
-  "tag_base": "v0.3"
+  "phase_branch": "phase/02-messaging",
+  "tag_base": "v0.3",
+  "phase_tag": "v0.3.2",
+  "release_id": 483
 }
@@ -130,9 +130,9 @@ NFR milestone: no separate minor tag. The final patch (v0.2.6) IS the v0.3 deliv

 | Phase | Name | Type | Status | Key Deliverables |
 |-------|------|------|--------|------------------|
-| 0 | Pre-Execution | docs | active | Spec, clarify, research, ideate, plan, PERSONAS.md (adds edge-engineer + languages-engineer phase-specific personas) |
-| 1 | Edge Domain | docs | pending | domains/edge/{first-principles, cdn, offline-first, iot, sync}.md |
-| 2 | Messaging Domain | docs | pending | domains/messaging/{first-principles, queues, pubsub, streams, delivery-semantics}.md |
+| 0 | Pre-Execution | docs | complete | Spec, clarify, research, ideate, plan, PERSONAS.md (adds edge-engineer + languages-engineer phase-specific personas) — shipped v0.3.0 |
+| 1 | Edge Domain | docs | complete | domains/edge/{first-principles, cdn, offline-first, iot, sync}.md — shipped v0.3.1 |
+| 2 | Messaging Domain | docs | complete | domains/messaging/{first-principles, queues, pubsub, streams, delivery-semantics}.md — shipped v0.3.2 |
 | 3 | Language-Derived Extension | docs | pending | languages/ × 4 → first-principles + 4 derived docs each (16 derived docs) |
 | 4 | Matrix + Review Integration | docs | pending | matrix/principles-matrix.md (+20 mappings, 170→190), matrix/domain-coverage.md (+ languages/ sub-table), review/{agent-checklist, peer-review-checklist, anti-patterns}.md, MANIFEST.md (languages/ section) |
 | 5 | Examples + Cross-Links | docs | pending | examples/good + examples/bad for 2 domains, cross-links to devops/observability/data/concurrency/kubernetes/infrastructure-as-code + language→domain links |
@@ -0,0 +1,227 @@
+# CDN — Derived Rules
+
+> Derives from `domains/edge/first-principles.md`. Applies P1
+> (Proximity is the Design Driver) and P6 (Cache Invalidation is
+> Explicit) primarily, with P5 (idempotent cache fill), P8
+> (geographic distribution), and P9 (identity at the edge). For the
+> edge-cache-vs-origin decision, see the decision matrix below.
+> Cross-links `domains/performance/frontend` for generic caching,
+> `domains/security/input-validation` for cache poisoning, and
+> `domains/observability/metrics` for cache-hit ratio.
+
+## What a CDN Is (P1 Proximity is the Design Driver)
+
+- A content delivery network is a fleet of PoPs (points of presence)
+  placed near users. The PoP serves cached content; the origin is
+  the authoritative source. The CDN's whole purpose is P1: compute
+  (the cache) is placed near the user so the round trip to the origin
+  does not bound latency. Latency is a correctness constraint at the
+  edge (C1), not a performance preference.
+- The CDN is the canonical edge-cache architecture (Akamai,
+  Cloudflare, Fastly): PoPs near users, origin shielding, cache-key
+  normalization, purge APIs. Atelier derives the
+  placement/invalidation principles, not the vendor config.
+- The boundary with `domains/performance/frontend` is per D-061:
+  performance owns *generic* caching and optimization (cache what is
+  expensive, stable, read often — `performance/P5 Caching with
+  Intent`); edge owns the *geographic, partition-aware* placement and
+  invalidation angle. A CDN is an edge concern because its defining
+  trait is geographic distribution (P8) and partition-aware
+  invalidation (P6), not measurement.
+
+## Cache Key Design (P6 Cache Invalidation is Explicit)
+
+- The cache key is the contract between the URL and the cached
+  representation. A key that varies on the wrong dimensions serves
+  the wrong content; a key that varies on too many dimensions
+  collapses the hit ratio. Key design *is* the invalidation
+  surface: a key that includes a content hash or version segment
+  makes invalidation explicit; a key that ignores `Vary` headers
+  serves stale variants.
+- Normalize the key: lower-case the host, strip default ports,
+  sort query parameters, ignore tracking parameters. A
+  non-normalized key is a cache-poisoning vector (see
+  `domains/security/input-validation`) and a hit-ratio destroyer
+  (see `domains/observability/metrics`).
+- A cache with no explicit key strategy is a TTL-less cache under
+  partition (P6 violation): staleness is silent and unbounded.
+
+```http
+# Cache key derivation: vary on what changes content, ignore what
+# does not. The key is the tuple (host, normalized-path, sorted-
+# query, Vary-headers); the cache entry is the representation + TTL.
+Cache-Key: example.com /api/v1/products?sort=price&region=us Vary:Accept-Encoding
+Cache-Control: public, max-age=60, s-maxage=600, stale-while-revalidate=300
+Vary: Accept-Encoding
+```
+
+- `max-age` bounds the browser cache; `s-maxage` bounds the CDN
+  PoP; `stale-while-revalidate` allows serving stale while
+  refetching. Each is an explicit invalidation strategy (P6).
+
+## TTL vs Explicit Invalidation (P6, C3 Simplicity)
+
+- **TTL-based invalidation** (`max-age`, `s-maxage`): the cache entry
+  expires after a duration. Simple, no origin contact required to
+  invalidate, but bounded staleness is the contract — the entry may
+  be stale up to TTL. Fits content where eventual consistency is
+  acceptable (asset fingerprints, lists, derived images).
+- **Explicit invalidation** (purge, surrogate keys): the operator
+  signals the cache to drop entries. Tighter staleness bounds, but
+  requires the origin or operator to know which entries to purge.
+  Fits content where staleness is a correctness defect (price
+  updates, availability, breaking news).
+- A TTL-less cache with no explicit invalidation is a P6 violation:
+  stale-forever under partition. Every cache must have one or the
+  other (or both), and the choice is documented per content type.
+
+## Cache-Hit / Miss / Origin-Fetch (P5, P6)
+
+- **Hit**: the PoP serves from cache. Latency is PoP-local (P1).
+- **Miss**: the PoP has no entry; it fetches from the origin (or an
+  origin-shield PoP). The fetch must be idempotent (P5) — a retried
+  miss must not corrupt the cache or double-write side effects.
+- **Revalidate**: the PoP holds a stale entry and asks the origin
+  (`If-None-Match`, `If-Modified-Since`); a 304 refreshes the TTL
+  without re-fetching the body. Revalidation is the bandwidth-economical
+  middle ground (C8).
+
+```http
+# Conditional revalidation — the PoP asks the origin "is this still
+# current?" The 304 response refreshes the TTL without a body.
+GET /api/v1/products HTTP/1.1
+Host: example.com
+If-None-Match: "etag-7a3f"
+
+HTTP/1.1 304 Not Modified
+ETag: "etag-7a3f"
+Cache-Control: s-maxage=600
+```
+
+- A cache-hit ratio that is not measured is a gate on noise — see
+  `domains/observability/metrics` for the SLI/SLO discipline that
+  makes the hit ratio a meaningful signal. A CDN with no hit-ratio
+  metric is operating blind (P10 analog).
+
+## Origin Shielding (P1, P8, C8 Economy)
+
+- Origin shielding routes all origin fetches through a single
+  shield PoP (or shield region). The shield absorbs the
+  thundering-herd: 10 000 PoPs missing the same URL fetch the origin
+  once, not 10 000 times. This is C8 Economy (origin bandwidth is
+  bounded) and P1 (the shield is itself a proximity layer for the
+  origin).
+- Shielding is a geographic decision (P8): the shield sits in a
+  region close to the origin, not close to the user. The shield is
+  the inner ring of the CDN; the user-facing PoPs are the outer ring.
+- A CDN without origin shielding under a stampede will overload the
+  origin; a shield that is itself partitioned from the origin must
+  degrade gracefully (P7) — serve stale per `stale-while-revalidate`
+  rather than 500.
+
+## Purge Strategies (P6, C3 Simplicity)
+
+| Strategy | Granularity | Latency to Invalidate | Cost | Best for |
+|----------|-------------|-----------------------|------|----------|
+| URL purge | One URL | Seconds | Low (one entry) | Surgical fixes, single-page corrections |
+| Soft purge | One URL (mark stale, serve while refetch) | Seconds | Low | High-traffic URLs where a hard purge causes a stampede |
+| Surrogate-key purge | A tag set (e.g., `product:123`, `category:shoes`) | Seconds | Medium (key indexing) | Related-content invalidation (a product update purges all its category pages) |
+| Wildcard purge | A path prefix or pattern | Seconds to minutes | High (scan) | Site-wide template changes |
+| All-cache purge | Everything | Seconds | Very high (origin stampede) | Disaster recovery only; never the steady-state invalidation path |
+
+- Surrogate-key purge (Fastly, Akamai) is the highest-value
+  strategy: tag cache entries with content keys, then purge by tag.
+  This is explicit invalidation at scale (P6) without the origin
+  stampede of an all-cache purge.
+- An all-cache purge as the steady-state invalidation path is a P6
+  violation dressed as a feature — it pushes the origin load back to
+  100% miss, defeating the CDN's purpose (P1).
+
+## Cache Poisoning Prevention (P9, cross-link security/input-validation)
+
+- A cache poisoned by a crafted request (a URL with a malicious
+  header that gets cached and served to others) is a correctness
+  defect (C1) and a security breach (P9 — the edge node is
+  exploited). Prevent poisoning by:
+  - Normalizing the cache key (strip untrusted query parameters,
+    ignore unknown headers, lower-case the host).
+  - Validating `Vary` against an allow-list; never `Vary: *` on a
+    shared cache (poisonable via header injection).
+  - Treating uncacheable responses (`Set-Cookie`,
+    `Cache-Control: private`) as never-stored.
+- See `domains/security/input-validation` for the general
+  input-validation discipline the cache key must follow. The cache
+  key is a validation surface; a non-validated key is an attack
+  surface.
+
+## Multi-CDN Routing (P8 Geographic Distribution)
+
+- A multi-CDN strategy routes each request to the best PoP across
+  providers (Akamai + Cloudflare + Fastly). Routing is
+  location-aware (P8): latency, cost, and availability vary by
+  region and provider. The DNS layer (or a client-side router)
+  selects the CDN per request.
+- Multi-CDN is a P8 decision, not a vendor-management decision:
+  geographic distribution is the first-class constraint. A
+  single-CDN deployment routes everything to one provider's PoPs;
+  a multi-CDN deployment routes by region, latency, and cost.
+- Invalidation across multiple CDNs is harder (P6): each provider
+  has its own purge API and surrogate-key scheme. A multi-CDN purge
+  must fan out to all providers; a purge that reaches only one CDN
+  leaves the others stale. Track purge completion per provider —
+  see `domains/observability/metrics` for the per-CDN hit-ratio and
+  purge-latency signals.
+
+```http
+# A CDN config example: cache-control headers + a purge rule.
+# Origin response: declare the cache contract (P6).
+HTTP/1.1 200 OK
+Cache-Control: public, max-age=60, s-maxage=600, stale-while-revalidate=300
+Surrogate-Key: product:123 category:shoes
+ETag: "etag-7a3f"
+Vary: Accept-Encoding
+
+# Purge rule (Fastly-style surrogate-key): when product 123
+# updates, purge every cache entry tagged product:123 OR
+# category:shoes. Explicit, bounded, no origin stampede (P6).
+POST /service/svc1/purge
+Surrogate-Key: product:123 category:shoes
+# Returns: {"status": "ok", "id": "purge-abc"} — poll the purge
+# status to confirm completion across all PoPs (P8, P10).
+```
+
+## Edge-Cache vs Origin — Decision Matrix (D-069)
+
+| Strategy | When | Latency | Origin Load | Correctness Risk |
+|----------|------|---------|-------------|------------------|
+| Serve from PoP (cache hit) | The PoP holds a fresh entry (within TTL or revalidated) | Lowest (PoP-local, P1) | None | Low — bounded by TTL staleness (P6) |
+| Serve stale while revalidate | The PoP holds a stale entry and `stale-while-revalidate` is set | Low (stale served immediately, refetch in background) | Background refetch (1 per entry) | Medium — stale served up to the revalidate window; acceptable for eventually-consistent content |
+| Fetch fresh from origin (miss) | The PoP has no entry, or the content is non-cacheable | High (origin round trip) | Full fetch per miss | Low — fresh by construction; the miss is the correctness floor |
+| Origin-shield fetch | Multiple PoPs miss the same URL; the shield collapses the herd | Medium (PoP → shield → origin) | Bounded to one origin fetch per shield (C8) | Low — shield is the inner ring; staleness bounded by shield TTL |
+| Purge and serve fresh | Explicit invalidation received (surrogate-key or URL purge) | Medium (purge propagates, then fresh fetch) | Full fetch post-purge | Lowest — explicit invalidation is the tightest staleness bound (P6) |
+| Serve from origin directly (bypass cache) | Content is non-cacheable (personalized, real-time) | Highest (every request hits origin) | Full fetch per request | Lowest for correctness, highest for origin load — use sparingly |
+
+- The default is **serve from PoP** when fresh, **fetch fresh from
+  origin** on miss with **origin-shield** to bound origin load, and
+  **purge and serve fresh** when explicit invalidation is required.
+  Bypass-the-cache is for non-cacheable content only — bypassing for
+  cacheable content is a P1 violation (you have defeated the CDN).
+- The correctness risk column is bounded by the invalidation
+  strategy (P6): every row except "bypass" carries staleness risk
+  that is bounded by TTL or explicit purge. A row with no
+  invalidation strategy is a P6 violation.
+
+## What Violates CDN Discipline
+
+| Violation | Principle |
+|-----------|-----------|
+| TTL-less edge cache under partition (stale-forever, no explicit invalidation) | P6 Cache Invalidation is Explicit |
+| Cache key that varies on untrusted query parameters (poisonable) | P6, P9 (`domains/security/input-validation`) |
+| All-cache purge as the steady-state invalidation path (origin stampede) | P6, C8 Economy |
+| Non-normalized cache key (case-sensitive host, unsorted query) | P6, `domains/security/input-validation` |
+| Bypass-the-cache for cacheable content | P1 Proximity is the Design Driver (defeats the CDN) |
+| Multi-CDN with no per-CDN purge completion tracking | P8, P10 (stale cache invisible to the operator) |
+| Origin fetch that is not idempotent under retry | P5 Edge Operations are Idempotent |
+| Cache-hit ratio not measured | P10, `domains/observability/metrics` |
+| Single-CDN deployed where geographic distribution requires multi-CDN | P8 Geographic Distribution |
+| Shield PoP that 500s instead of serving stale under partition | P7 Partial Degradation is Engineered |
@@ -0,0 +1,218 @@
+# Edge — First Principles
+
+## 1. The Principles
+
+### P1. Proximity is the Design Driver
+Compute, storage, and data are placed near the user or the data
+source. At the edge, latency is a correctness constraint (C1), not a
+performance preference — a late answer is a wrong answer when the
+round trip to a central region exceeds the user's or device's
+tolerance. This is the geographic expression of `C4 Locality`:
+performance's locality is algorithmic (data near compute); edge's
+locality is geographic (compute near user/data source). Placement is
+a design decision, not an accident of deployment, and it is
+constrained by `P8 Geographic Distribution`. The proximity angle is
+the distinguishing trait of the edge domain per D-061: this is what
+separates edge from `domains/performance/` (which owns *generic*
+measurement and optimization, not placement).
+
+### P2. Offline is a First-Class State
+The system continues to operate when disconnected from the center.
+Partition is the norm, not the exception; reconciliation happens on
+reconnect, never assumed to be instant. An app that crashes on
+disconnect has no offline state and is unengineered. Offline
+operation derives from `C5 Reversibility` — the disconnected state
+is reversible back to consistency via reconciliation — and `C1
+Correctness`, because correctness under partition is the contract,
+not eventual correctness as a hedge. This is the foundation for
+`domains/edge/offline-first.md` and the precondition for the
+bounded-conflict discipline of `P4`.
+
+### P3. Resources are Constrained and Declared
+Edge nodes — IoT sensors, gateways, point-of-sale devices, CDN PoP
+caches, 5G MEC nodes — have bounded CPU, memory, power, and
+bandwidth. Constraints are declared per node class, never assumed
+infinite. An undeclared budget is a defect: unbounded growth is a
+bug, and a constrained device with no budget will OOM or exhaust
+power. This derives from `C8 Economy` (use no more than the task
+requires) and `C1 Correctness` (a node that exceeds its bounds has
+failed). This is the edge-specific angle on `domains/performance/P4
+Resource Bounds` — performance owns the generic principle; edge owns
+the constrained-device reality. See `domains/edge/iot.md` for the
+per-device-class application.
+
+### P4. Sync Conflicts are Bounded, Not Infinite
+Divergent state across partitioned nodes converges. Oscillation and
+infinite sync loops are correctness failures, not eventual
+consistency. A merge that never terminates is a livelock; a CRDT
+without merge semantics or an LWW without a monotonic clock can
+oscillate forever. This derives from `C1 Correctness` (convergence is
+a correctness contract) and `C5 Reversibility` (divergent state is
+reversible back to convergence). The bound may be eventual (CRDTs) or
+arbitrated (LWW with vector clocks), but it must exist. This is the
+foundation for `domains/edge/sync.md` and the rule the
+`edge-sync-loop` chaos anti-pattern breaches.
+
+### P5. Edge Operations are Idempotent
+Sync, cache fill, and device commands are retried by nature — the
+network is partition-prone and the operation will be re-attempted.
+Idempotency keys (or deterministic operations) make retries safe. A
+non-idempotent edge write retried with side effects doubles the
+effect; a non-idempotent cache fill under retry corrupts the cache.
+This derives from `C1 Correctness`: correctness under retry is the
+contract, not a nice-to-have. This parallels `domains/messaging/P3
+Consumers are Idempotent` (cross-process delivery) and is the edge's
+device-and-cache-flavored analog — see `domains/edge/iot.md` for
+device command idempotency and `domains/edge/cdn.md` for cache-fill
+idempotency.
+
+### P6. Cache Invalidation is Explicit
+Edge caches carry a defined invalidation or TTL strategy. A
+stale-forever cache under partition is a silent correctness defect;
+a TTL-less cache with no explicit invalidation is a bug, not a
+feature. This derives from `C1 Correctness` (cached state must be
+correct) and `C3 Simplicity` (a defined invalidation strategy is
+simpler and clearer than ad-hoc staleness). This is distinct from
+`domains/performance/P5 Caching with Intent`, which owns *generic*
+caching and optimization; edge owns the *geographic,
+partition-aware* invalidation angle — when a PoP is partitioned from
+the origin, the invalidation strategy is the correctness mechanism.
+See `domains/edge/cdn.md` for purge strategies (URL vs soft vs
+surrogate-key) and the edge-cache-vs-origin decision matrix.
+
+### P7. Partial Degradation is Engineered
+The system degrades gracefully when an edge node or link fails. A
+partial service is a designed state with a defined contract, not a
+crash. One node's failure must not collapse the whole fleet; the
+degraded mode is documented, observable, and recoverable. This
+derives from `C1 Correctness` (the degraded contract is a
+correctness bound) and `C5 Reversibility` (recovery from degradation
+is reversible by construction). A crash-on-node-failure system has
+no degradation contract — it has an all-or-nothing failure mode that
+violates the fleet assumption. See `domains/edge/iot.md` for
+device-drop degradation and `domains/edge/offline-first.md` for
+partition degradation.
+
+### P8. Geographic Distribution is a First-Class Constraint
+The fleet is geo-distributed; routing, fan-out, and data placement
+are location-aware decisions, not accidents of deployment. The
+system is many nodes across many locations, not a single deployment.
+Data residency, regional latency, and PoP selection are engineered,
+not discovered in production. This derives from `C4 Locality` (the
+placement of data and compute is a locality decision) and `C6
+Composability` (the fleet composes from location-aware parts, each
+with its own contract). This is the structural companion to `P1
+Proximity`: P1 says *where* compute should be (near the user); P8
+says the *distribution* of compute across geographies is a
+first-class constraint. See `domains/edge/cdn.md` for multi-CDN
+routing.
+
+### P9. Identity is Constrained at the Edge
+Edge devices and nodes hold scoped, minimal credentials. No edge
+node is a cluster-admin-equivalent; device identity is per-device,
+not shared. One compromise must not equal a fleet compromise. This
+derives from `C1 Correctness` (security is a subset of correctness —
+an exploitable edge node does not do what it was supposed to do) and
+`C8 Economy` of trust (the credential scope is minimal for the task).
+A shared edge-device credential is the edge analog of a
+cluster-admin GitOps robot — blast radius is unbounded. See
+`domains/security/secrets.md` for the general secret-hygiene
+principles and `domains/edge/iot.md` for device provisioning.
+
+### P10. Edge Observability Survives Partition
+Telemetry is local-first: buffered on the node and forwarded on
+reconnect. Partition does not blind the operator. A fire-and-forget
+telemetry pipeline loses data when the link drops; a local-first
+buffer survives. This derives from `C7 Observability` (the fleet's
+behavior is visible to the operator) and `C5 Reversibility` (the
+buffered telemetry is reversible back to visibility on reconnect).
+This is distinct from `domains/observability/P1 Structured by
+Default`, which owns *generic* structured telemetry; edge owns the
+*partition-survivable, local-first* angle. See
+`domains/observability/metrics.md` and
+`domains/observability/logging.md` for the generic structured-
+telemetry foundations edge builds on.
+
+## 2. Core Principle Trace
+
+Each edge P-rule derives from one or more core C-rules (C1–C8). The
+matrix extension lands in P4 of the v0.4 plan; the traces below are
+authoritative. Edge is a broad-derivation domain touching 7 of 8
+core principles (C1, C3, C4, C5, C6, C7, C8); C2 (Clarity) is not a
+primary derivation — edge clarity is indirect (a cache with explicit
+invalidation is clearer than one without, but the primary trace is
+C1/C3).
+
+| P-rule | Core | Why |
+|--------|------|-----|
+| P1 Proximity is the Design Driver | C4, C1 | Locality of compute near user/data; correctness via latency |
+| P2 Offline is a First-Class State | C1, C5 | Correctness under partition; reversibility of reconciliation |
+| P3 Resources are Constrained and Declared | C8, C1 | Economy of constrained nodes; correctness of declared bounds |
+| P4 Sync Conflicts are Bounded, Not Infinite | C1, C5 | Correctness of convergence; reversibility of divergent state |
+| P5 Edge Operations are Idempotent | C1 | Correctness under retry |
+| P6 Cache Invalidation is Explicit | C1, C3 | Correctness of cached state; simplicity of defined invalidation |
+| P7 Partial Degradation is Engineered | C1, C5 | Correctness of degraded modes; reversibility of recovery |
+| P8 Geographic Distribution is a First-Class Constraint | C4, C6 | Locality of placement; composability of the fleet |
+| P9 Identity is Constrained at the Edge | C1, C8 | Correctness via security; economy of trust |
+| P10 Edge Observability Survives Partition | C7, C5 | Observability of the fleet; reversibility of buffered telemetry |
+
+## 3. What Violates These Principles
+
+| Violation | Principle Breached |
+|-----------|-------------------|
+| Central-region-only deployment for a latency-bound workload | P1 Proximity is the Design Driver |
+| App that crashes on disconnect (no offline state) | P2 Offline is a First-Class State |
+| Undeclared edge-node resource budget (assumes infinite CPU/memory) | P3 Resources are Constrained and Declared |
+| Sync loop that oscillates forever (CRDT without merge-semantics, LWW without monotonic clock) | P4 Sync Conflicts are Bounded, Not Infinite |
+| Non-idempotent edge write (cache-fill or device command retried with side effects) | P5 Edge Operations are Idempotent |
+| TTL-less edge cache under partition (stale-forever, no explicit invalidation) | P6 Cache Invalidation is Explicit |
+| Crash-on-node-failure (no partial-degradation contract) | P7 Partial Degradation is Engineered |
+| Random geographic placement (no location-aware routing) | P8 Geographic Distribution is a First-Class Constraint |
+| Shared edge-device credential (one key for the whole fleet) | P9 Identity is Constrained at the Edge |
+| Fire-and-forget telemetry (no on-node buffer; data lost on partition) | P10 Edge Observability Survives Partition |
+| Blocking call on a constrained IoT device with no timeout | P3, P5 (blocks the node; retry unsafe without idempotency) |
+| Multi-CDN routing with no PoP-selection logic (latency uncontrolled) | P8, P1 (placement not a design decision) |
+
+## 4. Relationship to Other Domains
+
+Edge computing is the engineering discipline of placing compute,
+storage, and data **near the source of generation or consumption**
+rather than in a centralized cloud. The distinguishing constraints are
+latency-bound operation, resource-constrained nodes,
+geo-distribution as a fleet, and partition-prone operation. Edge
+overlaps three existing domains by *subject* but not by *angle*: per
+D-061, edge owns the proximity/location/constraint/disconnection
+concerns that only arise at the network edge. The C4 Locality
+emphasis is the discriminator: performance's locality is algorithmic
+(data near compute); edge's locality is geographic (compute near
+user/data source). Cross-links are one-directional outward (per
+D-026 extended); no back-link edits to v0.1/v0.2/v0.3 content.
+
+- `domains/performance/frontend` ← P6 (edge owns geographic,
+  partition-aware cache invalidation; performance owns *generic*
+  caching and measurement — D-061 boundary)
+- `domains/performance/P4 Resource Bounds` ← P3 (edge owns
+  constrained-device reality; performance owns the generic
+  unbounded-growth-is-a-bug principle)
+- `domains/observability/metrics` ← P10 (cache-hit ratio, edge
+  telemetry aggregation; edge owns the local-first angle, observability
+  owns generic structured metrics)
+- `domains/observability/logging` ← P10 (local-first logging buffered
+  on-node and forwarded on reconnect)
+- `domains/concurrency/patterns` ← P5 (the offline write-queue is
+  the cross-partition analog of the in-process bounded buffer —
+  concurrency owns in-process; edge owns partition-survivable)
+- `domains/security/secrets` ← P9 (device credentials are scoped,
+  per-device, never shared — edge owns the constrained-identity
+  angle; security owns the general secret hygiene)
+- `domains/security/input-validation` ← P6 (cache poisoning
+  prevention — edge cache keys are a validation surface)
+- `domains/data/migrations` ← P4 (schema migration under sync must
+  reconcile across partitioned nodes; data owns the generic migration
+  discipline, edge owns the partitioned-reconcile angle)
+
+> Note: cross-links to `domains/messaging/` (e.g., MQTT QoS parallels
+> for delivery semantics) are intentionally omitted here — the
+> messaging domain is authored in P2. The intra-v0.4 edge↔messaging
+> links are added in P5 (ATELIER-114 per IDEATE-40) once both
+> domains exist; the dangling link is acceptable per D-053.
@@ -0,0 +1,258 @@
+# IoT — Derived Rules
+
+> Derives from `domains/edge/first-principles.md`. Applies P3
+> (Resources are Constrained and Declared) primarily, with P5
+> (command idempotency), P7 (partial degradation when devices drop),
+> P9 (device identity and provisioning), and P10 (telemetry from
+> devices). Cross-links `domains/security/secrets` for device
+> credentials and `domains/messaging/queues` for the MQTT QoS
+> parallels to delivery semantics.
+
+## What IoT at the Edge Is (P3 Resources are Constrained and Declared)
+
+- IoT at the edge is the engineering discipline of operating
+  constrained devices — sensors, actuators, gateways, microcontrollers
+  — as first-class participants in a distributed system. The
+  distinguishing constraint is per-device resource bounds (P3): a
+  battery-powered sensor has kilobytes of RAM, a constrained
+  protocol, and a multi-year sleep budget. These constraints are
+  declared per device class, never assumed infinite.
+- The boundary is per D-061: edge owns the constrained-device
+  reality; performance owns the generic unbounded-growth-is-a-bug
+  principle (`performance/P4 Resource Bounds`); concurrency owns
+  in-process primitives. IoT is an edge concern because its defining
+  traits are constrained resources (P3), geographic distribution as
+  a fleet (P8), partition-prone operation (P2), and device-scoped
+  identity (P9) — concerns that only arise at the network edge.
+- See `domains/edge/offline-first.md` for the partition-survival
+  discipline that constrained devices depend on, and
+  `domains/edge/sync.md` for the reconciliation of device state
+  across partitions.
+
+## Device Resource Classes (P3, C8 Economy)
+
+- A device resource class declares the bounds for a class of
+  devices: CPU (MHz, cores), memory (KB/MB), power (battery mAh,
+  duty-cycle budget), bandwidth (bytes/sec, latency budget), and
+  storage (KB/MB). Every device in the fleet is assigned to a class;
+  every operation is budgeted against its class.
+- An undeclared budget is a defect (P3 violation): a sensor that
+  sends telemetry every second without a duty-cycle budget exhausts
+  its battery in days, not years. The budget is the correctness
+  bound (C1) and the economy bound (C8).
+- A device class implies a protocol choice: a class-0 device
+  (constrained sensor, KB RAM) speaks CoAP; a class-1 device
+  (gateway, MB RAM) speaks MQTT; a class-2 device (edge compute
+  node, GB RAM) speaks HTTP. The protocol follows the constraint,
+  not the reverse.
+
+| Class | RAM | Power | Protocol | Typical role |
+|-------|-----|-------|----------|--------------|
+| 0 (constrained sensor) | < 10 KB | Battery, multi-year | CoAP, LoRaWAN | Telemetry only, no inbound commands |
+| 1 (actuator, gateway) | 10 KB – 1 MB | Battery or wired, weeks-months | MQTT, CoAP | Telemetry + commands, queue-and-forward |
+| 2 (edge compute) | > 1 MB | Wired, continuous | HTTP, MQTT | Local aggregation, gateway, edge inference |
+
+## Constrained Protocols — MQTT and CoAP (P3, P5)
+
+- **MQTT** is the canonical pub/sub protocol for constrained devices.
+  It is lightweight (2-byte header), broker-backed, and provides QoS
+  levels (0, 1, 2) that map to delivery semantics. MQTT is the
+  cross-process analog of message-queue delivery — see
+  `domains/messaging/queues` for the general queue/delivery-semantics
+  discipline; the cross-link is one-directional outward (edge →
+  messaging) per D-062 and D-026 extended.
+- **CoAP** is the REST analog for constrained devices: UDP-based,
+  low-overhead, with confirmable (CON) and non-confirmable (NON)
+  message types. CoAP fits class-0 devices where TCP is too heavy.
+- A blocking synchronous call on a constrained device with no
+  timeout is the `blocking-call-on-constrained-device` chaos
+  anti-pattern: it blocks the node, has no timeout (= hang), and
+  retries are unsafe without idempotency (P3 + P5 breach). Every
+  device operation must be async with a timeout, and every retried
+  operation must be idempotent.
+
+```json
+// MQTT publish/subscribe payload with QoS levels (P5 idempotency,
+// P3 constrained protocol).
+// QoS 0 — at-most-once: fire-and-forget, no ack. For telemetry
+// where a dropped sample is acceptable (P3 economy of the
+// constrained link).
+{
+  "topic": "devices/sensor-7/temperature",
+  "qos": 0,
+  "payload": {
+    "device": "sensor-7",
+    "ts": 1700000000,
+    "value": 21.4,
+    "unit": "C"
+  }
+}
+
+// QoS 1 — at-least-once: acked, may duplicate. The consumer must
+// be idempotent (P5) — dedup by (device, ts) or an idempotency key.
+{
+  "topic": "devices/actuator-3/command",
+  "qos": 1,
+  "payload": {
+    "device": "actuator-3",
+    "idempotencyKey": "cmd-1700000000-1",
+    "command": "set-point",
+    "value": 22.0
+  }
+}
+
+// QoS 2 — exactly-once: four-step handshake, no duplication. The
+// heaviest QoS; use only where duplicates are intolerable AND the
+// device has the budget for the handshake (class-1+ only, P3).
+{
+  "topic": "devices/actuator-3/irreversible-command",
+  "qos": 2,
+  "payload": {
+    "device": "actuator-3",
+    "idempotencyKey": "cmd-1700000000-2",
+    "command": "calibrate"
+  }
+}
+```
+
+- The QoS choice is a P5 (idempotency) and P3 (resource) decision:
+  QoS 0 is cheapest (no ack) but lossy; QoS 1 requires consumer
+  idempotency (P5); QoS 2 is exactly-once but costs a four-step
+  handshake on a constrained link. The default for telemetry is QoS
+  0 or 1; the default for commands is QoS 1 with an idempotency key
+  (P5); QoS 2 is reserved for irreversible commands where the
+  device budget permits.
+
+## Device Identity and Provisioning (P9 Identity is Constrained at the Edge)
+
+- Every device holds a per-device identity: a unique device ID and a
+  scoped credential (X.509 certificate, API token, or rotating
+  key). No shared fleet credential — one compromise must not equal
+  a fleet compromise (P9). The credential scope is minimal: a
+  device can publish to `devices/<its-id>/+` and subscribe to
+  `devices/<its-id>/commands`, nothing else.
+- Provisioning is the act of assigning a device identity at
+  enrollment time. The provisioning manifest declares the device,
+  its class, its allowed topics, and its credential. The manifest is
+  the P9 contract — a device operating outside its manifest scope is
+  a violation.
+- A device that is provisioned with a shared fleet key (the
+  `shared-edge-device-credential` anti-pattern) is a P9 violation:
+  blast radius is unbounded. See `domains/security/secrets` for the
+  general secret-hygiene discipline (per-identity credentials,
+  rotation, minimal scope) that device provisioning builds on.
+
+```yaml
+# Device provisioning manifest (P9 per-device identity + scoped
+# credentials). The manifest is the contract; the device operates
+# only within its declared scope.
+device:
+  id: sensor-7
+  class: 0                    # P3 resource class
+  model: temp-sensor-v2
+  firmware: 1.4.2
+identity:
+  cert: "sha256-of-device-cert"
+  credentialScope:
+    publish:
+      - "devices/sensor-7/temperature"
+      - "devices/sensor-7/status"
+    subscribe:
+      - "devices/sensor-7/commands"
+    # No wildcard, no fleet-wide topics (P9).
+provisioning:
+  enrolledAt: 2024-01-15T00:00:00Z
+  rotatesEvery: 90d
+  # Per-device credential; never shared (P9, domains/security/secrets).
+```
+
+## Telemetry from Devices (P10 Edge Observability Survives Partition)
+
+- Device telemetry is local-first (P10): the device buffers telemetry
+  on-node and forwards on reconnect. A fire-and-forget telemetry
+  pipeline loses data when the link drops; a buffered pipeline
+  survives. The buffer is bounded by the device class (P3): a
+  class-0 sensor buffers minutes of telemetry, not hours.
+- Telemetry is observable in aggregate: the operator sees the fleet's
+  behavior, not just per-device. A device that has not reported in
+  its expected interval is itself a signal (a dead device, a
+  partitioned device, a drained battery). See
+  `domains/observability/metrics` for the generic structured-metrics
+  discipline; edge owns the partition-survivable, local-first angle.
+- Telemetry must not be a secrets channel (P9 analog, see
+  `domains/observability/P6 No Secrets in Observability`): device
+  credentials, PII, and personally-identifying location must not
+  enter telemetry payloads.
+
+## Command Idempotency (P5 Edge Operations are Idempotent)
+
+- Device commands are retried by nature (the network is
+  partition-prone). Every command carries an idempotency key so a
+  retried command does not double-apply (P5). A `set-point` command
+  retried with the same idempotency key sets the point once, not
+  twice; an `open-valve` command retried is safe because the valve
+  is already open.
+- Irreversible commands (a calibration burn-in, a firmware flash)
+  require stronger idempotency: the device tracks applied
+  idempotency keys and refuses re-application. A retried irreversible
+  command without idempotency tracking double-applies the effect
+  (P5 violation, possibly a physical-side-effect bug).
+- The idempotency key is per-command, not per-device. A device that
+  dedups by device ID alone will drop distinct commands issued in
+  the same window. Use `(device, command-id, ts-window)` or a
+  UUID per command.
+
+## Partial Degradation When Devices Drop (P7 Partial Degradation is Engineered)
+
+- A fleet degrades when devices drop (battery exhaustion, partition,
+  hardware failure). The system must continue to operate with the
+  remaining devices; a whole-system crash on one device's failure is
+  a P7 violation. The degraded mode is documented: which
+  aggregations are valid with N-1 devices, which alerts fire, which
+  fallbacks engage.
+- A device that drops is not an incident by itself — fleets expect
+  churn. The operator-facing signal is the *aggregate* health (X%
+  of devices reporting, Y% partitioned for >Z minutes), not the
+  per-device drop. Per-device drop alerts are noise; aggregate
+  degradation alerts are signal (see `domains/observability/metrics`).
+- A command to a dropped device must time out (P5 — idempotent
+  retry) and degrade (P7 — the fleet continues without that
+  device). A command that blocks forever waiting for a dropped
+  device is the `blocking-call-on-constrained-device` chaos
+  anti-pattern (P3 + P5 breach).
+
+## Cross-Link to Messaging (P5, cross-link messaging/queues)
+
+- MQTT QoS 0/1/2 maps to at-most-once / at-least-once / exactly-once
+  delivery semantics — the same three-way tradeoff documented in
+  `domains/messaging/queues`. The cross-link is one-directional
+  outward (edge → messaging) per D-026 extended: edge owns the
+  constrained-device protocol angle; messaging owns the generic
+  cross-process delivery-semantics angle.
+- This link dangles until P2 (the messaging domain is authored in
+  P2); it is verified bidirectional in P5 (ATELIER-114 per
+  IDEATE-40). Acceptable per D-053 (vertical-slice integrity — P1
+  ships the edge domain self-consistent; the messaging cross-link
+  resolves by the P6 ship).
+- The parallel: a constrained device's QoS 1 publish is the
+  device-flavored instance of an at-least-once queue delivery — the
+  consumer (the broker or the downstream service) must be
+  idempotent (P5 here, `messaging/P3 Consumers are Idempotent`
+  there). The idempotency discipline is the same; the protocol and
+  failure model differ (constrained-device link vs broker-backed
+  network).
+
+## What Violates IoT-at-the-Edge Discipline
+
+| Violation | Principle |
+|-----------|-----------|
+| Undeclared device resource budget (assumes infinite battery/RAM) | P3 Resources are Constrained and Declared |
+| Shared fleet credential (one key for all devices) | P9 Identity is Constrained at the Edge |
+| Non-idempotent device command (retried command doubles the effect) | P5 Edge Operations are Idempotent |
+| Blocking synchronous call on a constrained device with no timeout | P3, P5 (blocks the node; retry unsafe) |
+| Fire-and-forget telemetry with no on-device buffer (lost on partition) | P10 Edge Observability Survives Partition |
+| Whole-system crash on one device's failure (no degradation contract) | P7 Partial Degradation is Engineered |
+| Device credential scope that includes fleet-wide topics (over-scoped) | P9, `domains/security/secrets` |
+| QoS 2 used on a class-0 device (no budget for the handshake) | P3 Resources are Constrained and Declared |
+| Per-device-drop alert (noise; aggregate degradation is the signal) | P7, `domains/observability/metrics` |
+| Telemetry payload that includes device credentials or PII | P9, `domains/observability/P6 No Secrets in Observability` |
@@ -0,0 +1,354 @@
+# Offline-First — Derived Rules
+
+> Derives from `domains/edge/first-principles.md`. Applies P2
+> (Offline is a First-Class State) primarily, with P5 (idempotent
+> queue-and-forward), P4 (bounded sync conflicts on reconnect), P7
+> (partial degradation), and P10 (local-first telemetry). Cross-links
+> `domains/concurrency/patterns` for the in-process bounded-buffer
+> analog and `domains/observability/logging` for local-first logging.
+
+## What Offline-First Is (P2 Offline is a First-Class State)
+
+- Offline-first is the design discipline in which the system
+  continues to operate when disconnected from the center. Partition
+  is the norm, not the exception; reconciliation happens on
+  reconnect. The offline state is engineered, not a degenerate mode
+  the app falls into by accident.
+- The boundary is per D-061: edge owns the
+  proximity/location/disconnection angle. An offline-first web app
+  is an edge concern because its defining trait is partition-survival
+  (P2), not generic performance. The local-first storage is the
+  edge device's constrained-resource reality (P3).
+- Offline-first is the precondition for the bounded-conflict
+  discipline of `P4 Sync Conflicts are Bounded, Not Infinite`:
+  without offline operation there is nothing to reconcile; with it,
+  the reconnect reconciliation is the correctness mechanism. See
+  `domains/edge/sync.md` for the conflict-resolution strategies.
+
+## Local-First Storage (P2, P3)
+
+- Local-first storage holds the working copy on the device:
+  IndexedDB (browser), SQLite (mobile, embedded), or on-device file
+  storage (desktop, IoT gateway). The local store is the authority
+  while offline; the server is reconciled later, not consulted per
+  read.
+- The local store is bounded by the device (P3 — Resources are
+  Constrained and Declared). A local store that grows without bound
+  is a defect: declare a budget (e.g., a 50 MB IndexedDB quota, a
+  30-day rolling window), and evict outside the budget deterministically.
+- The local store is the offline state; without it the app is
+  online-only and crashes on disconnect (P2 violation). The store is
+  the reversibility mechanism (C5): every local write is reversible
+  on reconcile.
+
+```typescript
+// Local-first store sketch (IndexedDB). The app reads from the
+// local store, never the network, while offline. Writes queue
+// locally and forward on reconnect (P2, P5).
+const db = await openDB("atelier-offline", 1, {
+  upgrade(db) {
+    const store = db.createObjectStore("pending-writes", {
+      keyPath: "id",
+    });
+    store.createIndex("by-createdAt", "createdAt");
+  },
+});
+
+async function readRecord(id: string) {
+  // Read from local store first; the network is a reconcile path,
+  // not the read path.
+  return db.get("pending-writes", id);
+}
+```
+
+## Queue-and-Forward for Writes (P5 Edge Operations are Idempotent)
+
+- Every write while offline is queued locally and forwarded to the
+  server on reconnect. The queue is the offline write-queue; the
+  forward is the reconcile. Each queued write carries an idempotency
+  key so a retried forward (the network is partition-prone) does not
+  double-apply (P5).
+- The queue is bounded (P3): a queue that grows without limit on a
+  constrained device will exhaust it. Declare a max-queue-depth and
+  a max-queue-bytes; reject or evict beyond the bound with a defined
+  policy (oldest-first, lowest-priority-first).
+- The queue is the cross-partition analog of the in-process bounded
+  buffer — see `domains/concurrency/patterns` (bounded buffer,
+  backpressure). Concurrency owns the in-process analog; edge owns
+  the partition-survivable analog. The failure model differs: the
+  in-process buffer fails by OOM; the offline write-queue fails by
+  partition or device loss.
+
+```typescript
+// Offline write-queue sketch. Each entry carries an idempotency
+// key (P5) so a retried forward is safe. The queue is bounded by
+// maxDepth (P3).
+interface PendingWrite {
+  id: string;            // local id
+  idempotencyKey: string; // server-side dedup key (P5)
+  collection: string;
+  payload: unknown;
+  createdAt: number;
+}
+
+const MAX_DEPTH = 1000;
+
+async function queueWrite(write: Omit<PendingWrite, "id" | "idempotencyKey" | "createdAt">) {
+  const depth = await db.count("pending-writes");
+  if (depth >= MAX_DEPTH) {
+    // P3: bounded queue. Evict the oldest pending write or reject.
+    // Rejecting is correct when the write is higher-priority than
+    // the oldest; evicting is correct when the newest is lowest.
+    throw new Error("offline-queue-full");
+  }
+  const entry: PendingWrite = {
+    ...write,
+    id: crypto.randomUUID(),
+    idempotencyKey: `${write.collection}:${crypto.randomUUID()}`,
+    createdAt: Date.now(),
+  };
+  await db.put("pending-writes", entry);
+  // The forward loop picks this up when connectivity returns.
+}
+
+async function forwardPendingWrites(server: Server) {
+  const pending = await db.getAllFromIndex("pending-writes", "by-createdAt");
+  for (const write of pending) {
+    // P5: idempotent — the server dedups by idempotencyKey.
+    await server.apply(write, write.idempotencyKey);
+    await db.delete("pending-writes", write.id);
+  }
+}
+```
+
+## Conflict Detection on Reconnect (P4 Sync Conflicts are Bounded)
+
+- On reconnect, the queued writes are forwarded; the server may
+  have advanced while the device was offline. A conflict is when the
+  local write and the server state diverge. Conflict detection is
+  the precondition for bounded reconciliation (P4): a write forwarded
+  blindly (last-write-wins with no clock) is a P4 violation waiting
+  to happen.
+- Conflict resolution strategies (CRDT, LWW with vector clocks,
+  application-specific merge) are the subject of
+  `domains/edge/sync.md` — the CRDT-vs-LWW decision matrix there
+  determines which applies. Offline-first owns the *detection*; sync
+  owns the *resolution*.
+- A reconnect that detects no conflicts when conflicts exist is a
+  silent correctness defect (C1, P4). Detection must be conservative:
+  when in doubt, flag a conflict and surface it to the merge
+  function or the user.
+
+## UI for Offline State (P7 Partial Degradation is Engineered)
+
+- The UI must reflect the offline state visibly: a "you are offline,
+  changes will sync when connected" banner, a pending-writes counter,
+  a last-synced timestamp. A UI that hides the offline state
+  violates P7 — the degraded mode is a designed state with a defined
+  contract, not a silent fall-through.
+- The UI must function while offline: reads from local-first
+  storage, writes to the queue, navigation that does not require the
+  network. An app that shows a blank screen or a spinner-forever when
+  offline has no offline state (P2 violation) and no degradation
+  contract (P7 violation).
+- The pending-writes counter is the local-first analog of the
+  messaging consumer-lag metric — see
+  `domains/observability/metrics` for the lag-discipline parallel.
+
+## Service Workers (P2, P6)
+
+- A service worker is a client-side proxy that intercepts network
+  requests and serves from a local cache. It is the browser's
+  offline-first primitive: the service worker cache is the
+  offline-capable store for assets; the IndexedDB store is the
+  offline-capable store for data.
+- The service worker cache is an edge cache (P6 — Cache Invalidation
+  is Explicit): it must carry a TTL or explicit invalidation
+  strategy. A service worker that caches forever and never
+  invalidates is a TTL-less edge cache under partition — a P6
+  violation (stale-forever).
+- See `domains/edge/cdn.md` for the generic edge-cache invalidation
+  discipline; the service worker is the on-device instance of it.
+
+```javascript
+// Service worker cache strategy: stale-while-revalidate for
+// assets, network-first for data, explicit version-bump for
+// breaking changes (P6).
+const CACHE = "atelier-v3";      // bump on deploy to invalidate (P6)
+const ASSETS = ["/", "/app.js", "/styles.css"];
+
+self.addEventListener("install", (event) => {
+  event.waitUntil(
+    caches.open(CACHE).then((cache) => cache.addAll(ASSETS))
+  );
+});
+
+self.addEventListener("fetch", (event) => {
+  const url = new URL(event.request.url);
+  if (url.pathname.startsWith("/api/")) {
+    // Network-first for data; fall back to cache on partition (P2).
+    event.respondWith(
+      fetch(event.request).catch(() => caches.match(event.request))
+    );
+  } else {
+    // Stale-while-revalidate for assets (P6 explicit invalidation).
+    event.respondWith(
+      caches.open(CACHE).then(async (cache) => {
+        const cached = await cache.match(event.request);
+        const network = fetch(event.request).then((resp) => {
+          cache.put(event.request, resp.clone());
+          return resp;
+        }).catch(() => cached);
+        return cached || network;
+      })
+    );
+  }
+});
+
+self.addEventListener("activate", (event) => {
+  // P6: explicit invalidation. Drop old caches on activate.
+  event.waitUntil(
+    caches.keys().then((keys) =>
+      Promise.all(keys.filter((k) => k !== CACHE).map((k) => caches.delete(k)))
+    )
+  );
+});
+```
+
+## Offline Write-Queue and Conflict Detection Mapped to the Testing Pyramid (IDEATE-38, ATELIER-94)
+
+The offline write-queue and conflict-detection patterns must be
+tested at every tier of the testing pyramid. Each tier exercises a
+different failure mode; skipping a tier leaves a correctness gap
+(P2, P4 violations that surface only in production partitions).
+
+| Pyramid Tier | What it exercises | What it proves |
+|-------------|-------------------|----------------|
+| **Unit** | Conflict detection on a merge function (pure inputs → expected merge result) | The merge logic is correct in isolation (P4) — given two divergent states, the merge returns the converged state |
+| **Integration** | Reconnect reconcile against a local store (fake server, real IndexedDB/SQLite) | The queue-and-forward loop drains correctly; the local store and server converge after reconnect (P2, P5) |
+| **E2e** | Partition simulation with a fake network (the app runs in a browser, the network is cut and restored) | The offline state, UI, and reconcile work end-to-end under partition (P2, P7) |
+
+- **Unit — conflict detection on a merge function.** The merge
+  function is pure: given two divergent states and a clock, it
+  returns the converged state. Test every merge case (LWW, CRDT
+  register, set union, application-specific three-way merge) as a
+  pure function. This is the cheapest tier and the highest coverage
+  per test — see `domains/testing/pyramid`.
+
+```typescript
+// Unit test sketch: conflict detection on a merge function (P4).
+// The merge function is pure; no network, no store. Test that
+// divergent states converge and that the merge is bounded (no
+// oscillation).
+
+function mergeLWW(local: State, remote: State, clock: Clock): State {
+  // Last-write-wins: the state with the later vector-clock wins.
+  // Returns the converged state (P4).
+  return clock.compare(local.clock, remote.clock) >= 0 ? local : remote;
+}
+
+// Unit cases:
+// - local ahead → local wins
+// - remote ahead → remote wins
+// - concurrent (clocks incomparable) → conflict flagged or LWW tiebreak
+// - identical → no-op convergence (bounded, no oscillation)
+test("mergeLWW converges when local is ahead", () => {
+  const local = { v: 2, clock: { a: 2 } };
+  const remote = { v: 1, clock: { a: 1 } };
+  expect(mergeLWW(local, remote, { compare: (a, b) => a.a - b.a })).toEqual(local);
+});
+```
+
+- **Integration — reconnect reconcile against a local store.** A
+  fake server stands in for the network; the real IndexedDB (or
+  SQLite) holds the queue. The test fills the queue while offline,
+  reconnects, and asserts the queue drains and the server and local
+  store converge. This exercises the queue-and-forward loop (P5)
+  and the reconcile against real storage.
+
+```typescript
+// Integration test sketch: reconnect reconcile against a local
+// store. A fake server; real IndexedDB. The queue drains; the
+// server and local store converge after reconnect (P2, P5).
+
+test("reconnect reconciles pending writes against the server", async () => {
+  const db = await openDB("test-offline", 1, { /* schema */ });
+  const server = new FakeServer();
+  await queueWrite(db, { collection: "docs", payload: { v: 1 } });
+  // Simulate offline: server is unreachable.
+  server.offline();
+  await queueWrite(db, { collection: "docs", payload: { v: 2 } });
+  expect(await db.count("pending-writes")).toBe(2);
+  // Simulate reconnect: server is reachable.
+  server.online();
+  await forwardPendingWrites(server, db);
+  expect(await db.count("pending-writes")).toBe(0);
+  expect(await server.latest("docs")).toEqual({ v: 2 });
+});
+```
+
+- **E2e — partition simulation with a fake network.** The app runs
+  in a real browser; a fake network layer cuts and restores the
+  connection. The test asserts the UI shows the offline state, the
+  writes queue, the reconnect reconciles, and the UI returns to
+  online. This is the highest-fidelity tier and the lowest coverage
+  per test — run a small number of representative scenarios, not a
+  combinatorial matrix.
+
+```typescript
+// E2e test sketch: partition simulation with a fake network. The
+// app runs in a browser; the network is cut and restored. Asserts
+// the offline UI state, the queue, the reconcile, and the online
+// recovery (P2, P7).
+
+test("app survives a network partition and reconciles on reconnect", async () => {
+  await page.goto("https://app.example.com");
+  await page.click("text=Edit document");
+  await page.fill("textarea", "offline edit");
+  // Cut the network.
+  await page.setOffline(true);
+  await page.click("text=Save");
+  await expect(page.locator("text=You are offline")).toBeVisible();
+  await expect(page.locator("text=1 pending change")).toBeVisible();
+  // Restore the network.
+  await page.setOffline(false);
+  await expect(page.locator("text=All changes synced")).toBeVisible();
+  await expect(page.locator("text=0 pending changes")).toBeVisible();
+});
+```
+
+- The three tiers are complementary: unit proves the merge logic,
+  integration proves the reconcile loop, e2e proves the partition
+  behavior. Skipping any tier leaves a correctness gap. See
+  `domains/testing/pyramid` for the pyramid discipline and
+  `domains/testing/fixtures` for the fake-server and fake-network
+  fixture patterns.
+
+## Observability (P10 Edge Observability Survives Partition)
+
+- The offline state is itself an observable signal: the
+  pending-writes count, the last-synced timestamp, the
+  reconcile-failure count. A device stuck offline for days with a
+  full queue is an incident; without local-first telemetry it is
+  invisible (P10 violation).
+- Local-first logging (buffered on-device, forwarded on reconnect)
+  is the offline-first instance of `P10 Edge Observability Survives
+  Partition`. See `domains/observability/logging` for the generic
+  structured-logging discipline the local-first buffer builds on.
+- A reconcile failure that is not logged locally is a silent defect
+  — the operator cannot debug what they cannot see (C7, P10).
+
+## What Violates Offline-First Discipline
+
+| Violation | Principle |
+|-----------|-----------|
+| App that crashes on disconnect (no offline state) | P2 Offline is a First-Class State |
+| Unbounded offline write-queue (grows until device OOM) | P3 Resources are Constrained and Declared |
+| Queued write forwarded without an idempotency key (retry doubles the effect) | P5 Edge Operations are Idempotent |
+| Reconnect that detects no conflicts when conflicts exist | P4 Sync Conflicts are Bounded, Not Infinite |
+| UI that hides the offline state (no banner, no pending counter) | P7 Partial Degradation is Engineered |
+| Service worker cache with no TTL and no explicit invalidation | P6 Cache Invalidation is Explicit |
+| Reconcile failure with no local log (silent under partition) | P10 Edge Observability Survives Partition |
+| Merge function that oscillates (no convergence guarantee) | P4, `domains/edge/sync.md` |
+| Local-first store with no declared budget (grows without bound) | P3, `domains/concurrency/patterns` (bounded buffer analog) |
+| E2e tests that never simulate a partition (offline path untested) | P2, `domains/testing/pyramid` |
@@ -0,0 +1,315 @@
+# Sync — Derived Rules
+
+> Derives from `domains/edge/first-principles.md`. Applies P4 (Sync
+> Conflicts are Bounded, Not Infinite) primarily, with P5
+> (idempotent merge operations), P2 (offline as the precondition),
+> and P10 (sync is observable). Cross-links `domains/data/migrations`
+> for schema migration under sync and `domains/concurrency/patterns`
+> for the immutability-aid-merge principle.
+
+## What the Sync Problem Is (P4 Sync Conflicts are Bounded, Not Infinite)
+
+- Sync is the discipline of reconciling divergent state across
+  partitioned nodes. While partitioned, each node accepts writes
+  independently; on reconnect, the divergent state must converge.
+  The correctness contract is that the merge terminates and
+  converges — oscillation and infinite sync loops are correctness
+  failures, not eventual consistency (P4).
+- Sync is the edge domain's deepest problem: it is the
+  reconciliation layer above `P2 Offline is a First-Class State`.
+  Without offline operation there is nothing to sync; with it, the
+  reconnect reconciliation is the correctness mechanism. See
+  `domains/edge/offline-first.md` for the offline write-queue that
+  produces the divergent state to be reconciled.
+- The boundary is per D-061: edge owns the partitioned-reconcile
+  angle; concurrency owns the in-process analog
+  (`concurrency/P1 Immutability by Default` — immutability aids
+  merge); data owns the generic migration discipline
+  (`data/migrations`). Sync is an edge concern because its defining
+  trait is partitioned divergence, a concern that only arises at the
+  network edge.
+
+## Conflict Resolution Strategies (P4, C1 Correctness, C5 Reversibility)
+
+- A conflict is when two nodes have divergent state for the same
+  logical entity and no total order determines which is correct.
+  Resolution strategies fall into two families:
+  - **Conflict-free**: the data type guarantees convergence by
+    construction (CRDTs). The merge is deterministic; no conflict
+    surfaces to the user or the application.
+  - **Conflict-tolerant**: the data type can conflict; the
+    resolution policy (last-write-win, three-way merge,
+    application-specific) arbitrates. Conflicts may surface to the
+    user or be silently resolved per a documented policy.
+- The choice is a P4 decision: conflict-free types guarantee the
+  bound (convergence) but constrain the data model; conflict-tolerant
+  types are flexible but require the resolution policy to be correct
+  and bounded (no oscillation). See the decision matrix below.
+
+## CRDTs — Conflict-Free Replicated Data Types (P4, C5, C6)
+
+- A CRDT is a data type whose merge operation is associative,
+    commutative, and idempotent. Given any set of divergent replicas,
+  merging them in any order converges to the same state — the merge
+  is deterministic and terminating (P4 bound). CRDTs derive from C5
+  Reversibility (divergent state reverses to convergence) and C6
+  Composability (CRDTs compose: a CRDT map of CRDT registers is
+  itself a CRDT).
+- **State-based (CvRDT — convergent):** each replica carries its
+  full state; merge is a least-upper-bound on a semi-lattice. The
+  payload is larger (full state per merge); the merge is simple
+  (one function). Fits small state and unreliable networks.
+- **Operation-based (CmRDT — commutative):** each replica carries
+  operations; merge is applying the operations in causal order. The
+  payload is smaller (ops, not state); the delivery must be
+  reliable and causally ordered. Fits large state and reliable
+  transport.
+- The tradeoff: state-based is simpler but heavier; operation-based
+  is lighter but requires causal delivery. Both guarantee
+  convergence (P4); the choice is a C8 Economy decision (bandwidth
+  vs delivery complexity).
+
+```typescript
+// CRDT register: LWW-element-set (state-based, CvRDT). The merge
+// is deterministic — the register with the later timestamp wins.
+// Convergence is guaranteed (P4); the merge is idempotent (P5).
+
+interface LWWRegister<T> {
+  value: T;
+  timestamp: number;   // monotonic clock; ties broken by node id
+  nodeId: string;
+}
+
+function mergeLWWRegister<T>(
+  local: LWWRegister<T>,
+  remote: LWWRegister<T>,
+): LWWRegister<T> {
+  // The merge is associative, commutative, idempotent (P4, P5).
+  // (local.timestamp, local.nodeId) > (remote.timestamp, remote.nodeId)
+  // is a total order — no oscillation, no infinite loop.
+  if (local.timestamp > remote.timestamp) return local;
+  if (local.timestamp < remote.timestamp) return remote;
+  // Tie: break by node id for a deterministic total order.
+  return local.nodeId > remote.nodeId ? local : remote;
+}
+
+// The register is a CRDT: merge(merge(a, b), c) === merge(a, merge(b, c))
+// for any replicas a, b, c. Convergence is guaranteed (P4).
+```
+
+```typescript
+// CRDT set: add-wins last-write-wins element set (state-based).
+// Each element carries a timestamp; remove only wins if the
+// remove-timestamp is later than the add-timestamp. This avoids
+// the remove-wins-vs-add race (P4) without surfacing a conflict.
+
+interface AWLWWSet<T> {
+  adds: Map<T, number>;     // element -> add-timestamp
+  removes: Map<T, number>;  // element -> remove-timestamp
+}
+
+function mergeAWLWWSet<T>(a: AWLWWSet<T>, b: AWLWWSet<T>): AWLWWSet<T> {
+  const adds = new Map<T, number>(a.adds);
+  const removes = new Map<T, number>(a.removes);
+  for (const [el, ts] of b.adds) {
+    adds.set(el, Math.max(adds.get(el) ?? 0, ts));      // add-wins union
+  }
+  for (const [el, ts] of b.removes) {
+    removes.set(el, Math.max(removes.get(el) ?? 0, ts));
+  }
+  return { adds, removes };
+}
+
+function contains<T>(set: AWLWWSet<T>, el: T): boolean {
+  const addTs = set.adds.get(el) ?? 0;
+  const rmTs = set.removes.get(el) ?? 0;
+  return addTs > rmTs;   // add wins on equal timestamp (P4 bounded)
+}
+```
+
+## Last-Write-Win (LWW) with Vector Clocks (P4, C1, C5)
+
+- LWW is the simplest conflict-tolerant strategy: the write with the
+  latest timestamp wins. It is cheap, but it silently discards
+  concurrent writes — the "lost update" is the correctness cost. LWW
+  is correct only when the timestamp is a total order (a monotonic
+  clock, not wall time), and when lost concurrent writes are
+  acceptable (e.g., caching, presence, ephemeral state).
+- **Vector clocks** are the timestamp that knows about concurrency.
+  A vector clock records the logical time of each node; two writes
+  are concurrent iff neither vector dominates the other. LWW with
+  vector clocks: a write that is causally later wins; a write that
+  is concurrent conflicts and is resolved by a tiebreak (node id,
+  wall time, or application policy).
+- The tiebreak is the P4 bound: the conflict must be resolved
+  deterministically (no oscillation) and the resolution must be
+  documented. A tiebreak by wall time alone (no vector clock) is a
+  P4 violation waiting to happen — wall time skews across nodes,
+  and a clock skew can flip the tiebreak, oscillating the merge.
+
+```typescript
+// LWW with vector clocks (conflict-tolerant, P4 bounded). The
+// vector clock records causal order; concurrent writes conflict;
+// the conflict is tiebroken deterministically (no oscillation).
+
+type VectorClock = Record<string, number>;  // nodeId -> counter
+
+function compareClock(a: VectorClock, b: VectorClock): "before" | "after" | "equal" | "concurrent" {
+  let aBefore = false, bBefore = false;
+  const keys = new Set([...Object.keys(a), ...Object.keys(b)]);
+  for (const k of keys) {
+    const av = a[k] ?? 0;
+    const bv = b[k] ?? 0;
+    if (av < bv) aBefore = true;
+    if (av > bv) bBefore = true;
+  }
+  if (aBefore && bBefore) return "concurrent";
+  if (aBefore) return "before";
+  if (bBefore) return "after";
+  return "equal";
+}
+
+interface LWWVectorState<T> {
+  value: T;
+  clock: VectorClock;
+  writerId: string;  // tiebreak: deterministic, no oscillation (P4)
+}
+
+function mergeLWWVector<T>(
+  local: LWWVectorState<T>,
+  remote: LWWVectorState<T>,
+): LWWVectorState<T> {
+  const order = compareClock(local.clock, remote.clock);
+  if (order === "before") return remote;       // remote causally later
+  if (order === "after" || order === "equal") return local;
+  // Concurrent: tiebreak by writer id (deterministic, P4 bounded).
+  return local.writerId > remote.writerId ? local : remote;
+}
+```
+
+- The merge is idempotent (P5): merging the same two replicas twice
+  yields the same result. The tiebreak by `writerId` is a total
+  order, so the merge cannot oscillate (P4 bound).
+- A vector-clock merge that surfaces the concurrent conflict to the
+  application (instead of tiebreaking) is also valid — the
+  application resolves per its own policy. The P4 bound is that the
+  resolution terminates; the policy determines whether the user sees
+  the conflict or the system silences it.
+
+## Merge Semantics (P4, P5, cross-link concurrency/patterns)
+
+- The merge function is the heart of sync. Its properties (P4):
+  - **Associative**: `merge(merge(a, b), c) === merge(a, merge(b, c))`.
+  - **Commutative**: `merge(a, b) === merge(b, a)`.
+  - **Idempotent**: `merge(a, a) === a` (P5 — retried merges are safe).
+- Immutability aids merge: an immutable state representation (the
+  CRDT payload, the LWW register with a clock) makes the merge a
+  pure function of two inputs, with no in-place mutation race. See
+  `domains/concurrency/patterns` (`concurrency/P1 Immutability by
+  Default`) for the in-process immutability principle; sync is the
+  cross-partition instance of it.
+- A merge that mutates in place is a P5 violation waiting to
+  happen: a retried merge mutates the same state twice, and the
+  result is not idempotent. Always merge into a new state; never
+  mutate the inputs.
+
+## Conflict-Free vs Conflict-Tolerant Data Types (P4, C3 Simplicity)
+
+- **Conflict-free (CRDTs):** the data type guarantees convergence.
+  The application never sees a conflict; the merge is deterministic.
+  The cost: the data model is constrained (counters, sets, registers,
+  maps of these). A conflict-free type for arbitrary JSON is hard;
+  a conflict-free type for a counter is a PN-counter.
+- **Conflict-tolerant (LWW, three-way merge, application policy):**
+  the data type can conflict; the resolution policy arbitrates. The
+  cost: the policy must be correct and bounded (no oscillation), and
+  the conflict may surface to the user. The benefit: any data model
+  can be made conflict-tolerant (just pick a tiebreak).
+- The choice is the decision matrix below. It is a P4 decision
+  (which bound), a C1 decision (which correctness cost is
+  acceptable), and a C3 decision (which simplicity is affordable).
+  See also `domains/data/migrations` for the schema-evolution angle
+  — a schema change under sync must be compatible with both
+  replicas, or the merge fails on the new shape.
+
+## Schema Migration Under Sync (P4, cross-link data/migrations)
+
+- A schema migration under sync is harder than a single-node
+  migration: both replicas must understand the new shape, or the
+  merge fails. The migration must be forward-and-backward compatible
+  across all replicas that may still hold the old shape — see
+  `domains/data/migrations` for the generic compatibility discipline.
+- A breaking schema change under sync requires a staged migration:
+  deploy the new-shape-aware merge first (it accepts both shapes),
+  then deploy the new shape, then deploy the old-shape-removing
+  merge. A big-bang schema change under sync is a P4 violation: the
+  replicas that have not yet upgraded will fail the merge, and the
+  sync will not converge.
+- The merge function's version awareness is the P4 bound: the merge
+  must handle every shape version that may exist in the fleet, or
+  reject (and surface) the merge rather than silently corrupting.
+
+## CRDT vs Last-Write-Win — Decision Matrix (D-069)
+
+| Strategy | When | Correctness Guarantee | Operational Cost | Failure Mode |
+|----------|------|------------------------|-------------------|--------------|
+| CRDT (state-based, CvRDT) | The data model fits a CRDT (counter, set, register, map of these); convergence must be guaranteed without surfacing conflicts; the network is unreliable (full-state merge tolerates dropped ops) | Strong eventual convergence — `merge(a, b) === merge(b, a)` for any replicas (P4 bound by construction) | Medium — full state per merge (bandwidth); semi-lattice merge function per type; CRDT library or hand-rolled | A bug in the merge function = silent divergence (C1); large state = bandwidth cost on constrained links (P3) |
+| CRDT (operation-based, CmRDT) | The data model fits a CRDT; bandwidth is constrained (ops are smaller than state); the transport is reliable and causally ordered | Strong eventual convergence — same guarantee, smaller payload | High — requires causal delivery (vector clock or broker with ordering); op transform must be idempotent (P5) | Causal-delivery violation = lost ops = divergence; op-transform bug = silent divergence |
+| Last-Write-Win (LWW) with vector clocks | The data model is arbitrary (any JSON, any record); concurrent writes are acceptable to discard or tiebreak; a total order tiebreak (node id) is acceptable | Bounded convergence — causally-later writes win; concurrent writes are tiebroken deterministically (P4 bound via tiebreak) | Low — simple merge (compare clocks, pick winner); no CRDT library; small payload | Concurrent writes are silently discarded (lost update); tiebreak by wall time = clock-skew oscillation (P4 violation); no vector clock = no concurrent-write detection = silent loss |
+| LWW with wall-clock timestamp only | The data model is ephemeral (cache, presence); lost updates are acceptable; the clock is roughly synchronized (NTP) | Weak — convergence eventually, but concurrent writes may oscillate with clock skew; no concurrent-write detection | Lowest — one timestamp per write; no clock vector | Clock skew = oscillation (P4 violation); concurrent writes silently lost; not a correctness-safe strategy for durable state |
+| Three-way merge (application-specific) | The data model is structured (documents, forms); conflicts should surface to the user or a domain-specific resolver; the merge is field-level | Bounded if the merge function is correct (associative, commutative, idempotent — P4, P5); conflicts surface per field | High — application-specific merge function per type; UI for conflict resolution; user-facing conflict surface | Merge-function bug = silent divergence or oscillation; unbounded conflict UI = user fatigue |
+
+- The default for structured state that must converge silently is a
+  **CRDT** (state-based for unreliable networks, operation-based for
+  bandwidth-constrained reliable transport). The default for
+  arbitrary JSON where lost concurrent updates are acceptable is
+  **LWW with vector clocks** (never wall-clock-only for durable
+  state). The default for user-facing documents where conflicts
+  should surface is **three-way merge** with a documented resolution
+  policy.
+- The failure-mode column is the P4 check: every row except
+  wall-clock-only LWW carries a bounded failure mode (the bug is in
+  the implementation, not the strategy). Wall-clock-only LWW carries
+  an unbounded failure mode (clock skew = oscillation) and is a P4
+  violation for durable state. Use it only for ephemeral state
+  where lost updates are acceptable.
+- The choice is a P4 decision (which bound) and a C1 decision
+  (which correctness cost). A CRDT guarantees convergence but
+  constrains the data model; LWW is flexible but discards concurrent
+  writes. Neither is universally correct; the matrix is the
+  decision tool.
+
+## Observability of Sync (P10 Edge Observability Survives Partition)
+
+- Sync is itself an observable operation: the merge count, the
+  conflict count, the convergence lag (time from reconnect to
+  convergence), and the divergent-replica count are first-class
+  signals. A sync that runs forever without converging is the
+  `edge-sync-loop` chaos anti-pattern (P4 breach); without
+  observability it is invisible until the user notices the stale
+  state.
+- A conflict that is silently resolved should be logged (the
+  resolution policy applied, the discarded write's idempotency key,
+  the winning write's clock). A conflict that surfaces to the user
+  should be metricated (the conflict rate, the resolution time).
+  See `domains/observability/metrics` for the generic discipline.
+- A divergent replica that has not converged after the expected
+  window is an incident; without a metric it is invisible (P10
+  breach). Wire sync convergence to an alert — the
+  `divergent-replica-count` is the sync analog of the messaging
+  `consumer-lag` metric.
+
+## What Violates Sync Discipline
+
+| Violation | Principle |
+|-----------|-----------|
+| Sync loop that oscillates forever (CRDT without merge-semantics, LWW without monotonic clock) | P4 Sync Conflicts are Bounded, Not Infinite |
+| LWW with wall-clock timestamp only on durable state (clock skew = oscillation) | P4, C1 (no concurrent-write detection) |
+| Merge function that mutates inputs in place (retried merge is not idempotent) | P5 Edge Operations are Idempotent, `domains/concurrency/patterns` |
+| Big-bang schema change under sync (replicas fail the merge) | P4, `domains/data/migrations` |
+| Conflict silently resolved with no log (the policy is invisible) | P10 Edge Observability Survives Partition |
+| Divergent replica with no convergence-lag metric (invisible stale state) | P10, `domains/observability/metrics` |
+| Three-way merge with an unbounded conflict UI (user fatigue, no termination) | P4 (the merge must terminate) |
+| Operation-based CRDT without causal delivery (lost ops = divergence) | P4, C1 (the delivery contract is the bound) |
+| Merge that surfaces every concurrent conflict to the user (no default policy) | P4, C3 (the default policy is the simplicity bound) |
+| Sync with no convergence test (the merge is untested under partition) | P4, `domains/testing/pyramid` |
@@ -0,0 +1,379 @@
+# Delivery Semantics — Derived Rules
+
+> Derives from `domains/messaging/first-principles.md`. Applies P4
+> (Delivery Semantics are Explicit), P3 (Consumers are Idempotent),
+> P2 (Ordering is a Property, Not an Assumption), P5 (Dead-Letter
+> Handling is Defined), and P6 (Backpressure is Bounded) primarily,
+> with P10 (DLQ depth as an alert). For the dead-letter strategy
+> decision, see the comparison table below. Includes a fenced
+> idempotency-key dedup-store example (IDEATE-39). Cross-links
+> `domains/concurrency/patterns` for the in-process retry/backoff
+> analog, `domains/errors/patterns` for errors as data for message
+> failures, and `domains/observability/metrics` for DLQ depth as an
+> alert.
+
+## The Three Delivery Semantics (P4 Delivery Semantics are Explicit)
+
+- **At-most-once**: a message is delivered 0 or 1 times; loss is
+  possible, duplication is not. The producer fires-and-forgets; the
+  broker does not ack; the consumer does not dedup. Lowest latency,
+  lowest implementation cost, lossy. Fits telemetry where a dropped
+  sample is acceptable (MQTT QoS 0, fire-and-forget metrics).
+- **At-least-once**: a message is delivered 1 or more times;
+  duplication is possible, loss is not. The producer sends and
+  waits for a broker ack; the consumer processes and acks; a crash
+  before the consumer's ack triggers redelivery. The consumer must
+  be idempotent (P3). The default for side-effecting operations
+  (orders, payments, commands). The vast majority of broker-backed
+  queues (SQS standard, RabbitMQ ack, MQTT QoS 1).
+- **Exactly-once**: a message is delivered exactly 1 time; no
+  loss, no duplication. In practice, this is at-least-once plus
+  idempotency (P3), or a transactional consume-process-produce loop
+  (P4 — see `domains/messaging/streams.md`). Jepsen analyses verify
+  broker claims: "exactly-once" requires independent verification;
+  the durable engineering practice is at-least-once with idempotent
+  consumers (P3), which collapses to exactly-once under correct
+  dedup.
+- The semantic is declared per channel (P4), not emergent. An
+  unstated semantic is a defect: the consumer guesses, and the
+  guess is wrong under the first failure. See
+  `domains/messaging/queues.md` for the three-semantics comparison
+  table (semantics, latency cost, implementation cost, when each
+  fits).
+
+## Idempotency (P3 Consumers are Idempotent)
+
+- Idempotency is the correctness property that makes at-least-once
+  safe. A consumer that processes the same message twice has the
+  same effect as processing it once. The mechanism is the
+  idempotency key: a per-message unique identifier the consumer
+  uses to dedup redeliveries.
+- The idempotency key is per-message, not per-producer or
+  per-session. A consumer that dedups by producer alone drops
+  distinct messages issued in the same window. Use a UUID per
+  message, or a deterministic key derived from the message content
+  (e.g., `(entity, operation, version)`).
+- The dedup store is bounded (P6 — Backpressure is Bounded): a
+  dedup store that grows without bound is a memory leak. Use a TTL
+  window longer than the broker's max-redelivery window, or a
+  bounded LRU. The TTL is the P6 bound: a key seen within the TTL
+  is a redelivery; a key older than the TTL is expired (the broker
+  has given up redelivering it).
+
+## Idempotency-Key Dedup Store (IDEATE-39, P3 — fenced example)
+
+The dedup store is the concrete mechanism that makes a consumer
+idempotent under at-least-once delivery. This is the fenced
+example required by IDEATE-39 (parallel to the v0.3 IDEATE-29
+signed-attestation fenced example): it is NOT prose-only — the
+consumer-with-dedup-store demonstrates P3 concretely.
+
+```python
+# Idempotent consumer with a dedup store (P3 Consumers are
+# Idempotent, P6 Backpressure is Bounded — the dedup store is
+# TTL-bounded). The consumer dedups by idempotency key before
+# processing; a redelivered message is a no-op, not a double-apply.
+
+import time
+
+# P6: the dedup store is bounded by a TTL window. The TTL must
+# exceed the broker's max-redelivery window; beyond the TTL, the
+# key is expired (the broker has given up). A dedup store with no
+# TTL is a memory leak (P6 violation).
+DEDUP_TTL_SECONDS = 24 * 3600   # longer than max-redelivery window
+
+class DedupStore:
+    """A TTL-bounded idempotency-key dedup store (P3, P6).
+
+    seen(key): True if the key was processed within the TTL.
+    mark(key): Record the key as processed (with a timestamp).
+    """
+    def __init__(self, backend):
+        # backend is a Redis, a DB, or an in-process LRU. The
+        # backend must be shared across consumer instances if the
+        # subscription is shared (P2/P3 — see domains/messaging/
+        # pubsub.md on shared vs independent subscriptions).
+        self.backend = backend
+
+    def seen(self, key: str) -> bool:
+        ts = self.backend.get(key)
+        if ts is None:
+            return False
+        if time.time() - ts > DEDUP_TTL_SECONDS:
+            # P6: expired. The broker has given up redelivering;
+            # this key is no longer a redelivery signal.
+            self.backend.delete(key)
+            return False
+        return True
+
+    def mark(self, key: str):
+        self.backend.set(key, time.time(), ttl=DEDUP_TTL_SECONDS)
+
+
+# The idempotent consumer: dedup before process, mark after
+# process, ack after mark. A crash before mark re-processes (the
+# dedup store does not have the key); a crash before ack
+# redelivers (the broker did not see the ack) and the dedup store
+# makes the redelivery a no-op (P3).
+def consume_idempotent(broker, dedup: DedupStore, process):
+    for message in broker.receive():
+        # P3: dedup BEFORE process. A redelivered message is a
+        # no-op, not a double-apply.
+        if dedup.seen(message["idempotencyKey"]):
+            broker.ack(message)        # already processed; skip
+            continue
+        try:
+            process(message["payload"])
+            # P3: mark AFTER process succeeds. A crash between
+            # process and mark re-processes (acceptable: the
+            # process must be idempotent OR the mark must be
+            # transactional with the process — see below).
+            dedup.mark(message["idempotencyKey"])
+            broker.ack(message)
+        except TransientError as exc:
+            # P6: bounded retry with backoff. Nack for redelivery;
+            # the broker redelivers after exponential backoff.
+            broker.nack(message, delay=backoff(message["attempt"]))
+        except PoisonError as exc:
+            # P5: poison message. Route to DLQ, do NOT retry
+            # forever (see the DLQ routing rule below).
+            route_to_dlq(broker, message, exc)
+            broker.ack(message)       # remove from the origin queue
+```
+
+- The order `process → mark → ack` gives at-least-once with
+  idempotent dedup: a crash before `mark` re-processes (the dedup
+  store does not have the key), and a crash before `ack`
+  redelivers (the dedup store makes the redelivery a no-op). If
+  `process` is not itself idempotent, the `process → mark` window
+  must be transactional (e.g., process and mark in one DB
+  transaction) — otherwise a crash in the window double-applies.
+- For a non-idempotent `process` (e.g., a payment that must not
+  double-charge), use a transactional dedup: process and mark in
+  one DB transaction, so the mark commits iff the process
+  commits. This is the "exactly-once via idempotency" pattern
+  (P4): at-least-once delivery plus a transactional
+  process-and-mark collapses to exactly-once under correct
+  transactional semantics.
+
+## Ordering (P2 Ordering is a Property, Not an Assumption)
+
+- The delivery semantic interacts with ordering (P2). At-least-once
+  with per-partition ordering: a redelivery within a partition
+  preserves order (the redelivered message re-appears in its
+  original position relative to other messages the consumer has
+  not yet seen). At-least-once with no ordering: a redelivery may
+  appear out of order relative to messages delivered after it.
+- The consumer must not assume an ordering property the broker
+  does not provide (P2). A standard queue delivers per-receive-node
+  arrival order but no global order and no order across redeliveries;
+  a FIFO queue delivers strict per-group order including across
+  redeliveries; a partitioned stream delivers strict per-partition
+  order across redeliveries. Document the property; do not assume
+  it. See `domains/messaging/queues.md` (FIFO vs standard) and
+  `domains/messaging/streams.md` (per-partition order).
+
+## Dead-Letter Strategies (P5 Dead-Letter Handling is Defined)
+
+- A poison message (unparseable, repeatedly failing, or exhausting
+  the retry budget) must be routed to a dead-letter queue, not
+  retried forever or silently dropped (P5). The DLQ is observable
+  (P10 — depth is an alert) and drainable (an operator can
+  inspect, replay, or discard with audit).
+- The dead-letter strategy determines when a message is
+  dead-lettered and what the operator sees. The choice is the
+  decision matrix below (D-069).
+
+## Dead-Letter Strategy Comparison (D-069)
+
+| Strategy | When It Applies | Failure Visibility | Operational Cost |
+|----------|-----------------|-------------------|------------------|
+| **Retry-count-limit** | A fixed max-redeliveries count (e.g., 5). After N redeliveries, route to DLQ. Simple, predictable. | The redelivery count is visible in the DLQ entry; the operator sees how many times it was retried. | Low — a counter per message; no backoff tuning. Risk: retries fire as fast as the broker redelivers, hammering a downstream that is already failing (P6 — no backoff = no backpressure escape). |
+| **TTL-with-backoff** | A max time-to-live for redelivery (e.g., 30 minutes) with exponential backoff between retries. After the TTL, route to DLQ. | The TTL and the backoff schedule are visible; the operator sees the retry timeline. | Medium — backoff tuning per message type. Benefit: backoff gives the downstream time to recover (P6 — the retry rate is bounded); fits transient failures (a downstream that is briefly unavailable). |
+| **Poison-queue** | A separate queue for messages that fail a specific check (unparseable, schema-invalid, unknown type) before any processing retry. Routed immediately, not retried. | The poison queue is a separate signal from the DLQ; the operator sees parse-vs-process failures distinctly. | Low — a routing rule per check. Benefit: distinguishes "never going to succeed" (poison) from "might succeed on retry" (DLQ). Use for unparseable messages that no retry will fix. |
+| **DLQ + alert** | Any of the above strategies, plus an alert on DLQ depth. The DLQ is observable (P10) — depth, age, and rate are alerted. | Highest — the operator is paged on DLQ growth; the DLQ is a first-class signal, not a graveyard. | Medium — alerting setup per DLQ. This is the P5/P10 floor: a DLQ without an alert is a silent correctness defect (poison messages accumulate invisibly). |
+
+- The default for transient failures is **TTL-with-backoff + DLQ
+  + alert**: backoff gives the downstream time to recover (P6),
+  the TTL bounds the retry budget (P5), the DLQ captures the
+  unprocessable, and the alert makes it visible (P10). The default
+  for unparseable messages is **poison-queue + alert**: route
+  immediately, do not retry a message no retry will fix.
+- **Retry-count-limit alone** (no backoff, no alert) is the
+  `messaging-unbounded-retry` chaos anti-pattern's cousin: it
+  caps the count but hammers the downstream at full retry rate,
+  and the DLQ grows silently if no alert is wired. Always pair
+  a retry budget with backoff (P6) and an alert (P10).
+- The failure-visibility column is the P10 check: every strategy
+  must surface the failure to the operator. A strategy with no
+  visibility is a P10 violation regardless of its retry semantics.
+- The operational-cost column is the C8 tradeoff: more visibility
+  and more backoff cost more to set up but pay back in operational
+  stability. The P5/P10 floor is "DLQ + alert"; below that, the
+  strategy is a silent defect waiting to grow.
+
+## DLQ Routing Rule (P5, P10)
+
+```python
+# DLQ routing rule (P5 dead-letter handling, P6 bounded retry with
+# backoff, P10 DLQ depth alert). Combines TTL-with-backoff for
+# transient failures and poison-queue for unparseable messages,
+# with an alert on DLQ depth.
+
+import json, time
+
+MAX_RETRY_TTL_SECONDS = 30 * 60    # 30 min total retry window
+DLQ = "orders-dlq"
+POISON = "orders-poison"           # unparseable; never retried
+
+def route_to_dlq(broker, message, reason):
+    """Route a message to the DLQ with audit metadata (P5)."""
+    broker.send(DLQ, body=json.dumps({
+        "original": message,
+        "reason": str(reason),
+        "deadLetteredAt": now_iso(),
+        "redeliveryCount": message.get("attempt", 0),
+    }))
+    # P10: emit a metric so DLQ depth alerts fire. A DLQ that
+    # grows with no alert is a silent correctness defect (P5/P10).
+    metrics.increment("dlq.depth", tags={"queue": "orders"})
+    # The ack removes the message from the origin queue; the DLQ
+    # is the durable record (P5 — observable and drainable).
+
+def consume_with_dlq(broker, dedup, process):
+    for message in broker.receive():
+        # P1: parse first. An unparseable message is poison —
+        # route immediately, do NOT retry (no retry will fix it).
+        try:
+            payload = json.loads(message["body"])
+        except (ValueError, SchemaError) as exc:
+            broker.send(POISON, body=json.dumps({
+                "original": message["body"],
+                "reason": f"parse-failed: {exc}",
+                "poisonedAt": now_iso(),
+            }))
+            metrics.increment("poison.depth", tags={"queue": "orders"})
+            broker.ack(message)       # remove from origin; poison queue holds it
+            continue
+
+        # P3: dedup before process.
+        if dedup.seen(payload["idempotencyKey"]):
+            broker.ack(message); continue
+
+        attempt = payload.get("attempt", 0)
+        first_attempt_ts = payload.get("firstAttemptTs", time.time())
+
+        try:
+            process(payload)
+            dedup.mark(payload["idempotencyKey"])
+            broker.ack(message)
+        except TransientError as exc:
+            # P6: TTL-with-backoff. If the retry window is
+            # exhausted, route to DLQ; otherwise redeliver with
+            # exponential backoff.
+            if time.time() - first_attempt_ts > MAX_RETRY_TTL_SECONDS:
+                route_to_dlq(broker, message, exc)   # P5
+                broker.ack(message)
+            else:
+                broker.nack(message, delay=backoff(attempt))
+        except PermanentError as exc:
+            # A permanent error (e.g., a not-found dependency)
+            # does not benefit from retry — route to DLQ now.
+            route_to_dlq(broker, message, exc)
+            broker.ack(message)
+```
+
+- The routing rule distinguishes three failure modes: **poison**
+  (unparseable — route immediately, no retry), **transient**
+  (retry with backoff until the TTL, then DLQ), and **permanent**
+  (a retry will not fix it — DLQ now). This distinction is the P5
+  discipline: not every failure is a retry; some are immediate
+  DLQs.
+- The DLQ entry carries `reason`, `deadLetteredAt`, and
+  `redeliveryCount` — it is auditable (the operator knows why each
+  message was dead-lettered and how many times it was retried).
+  This is the errors-as-data discipline — see
+  `domains/errors/patterns` for the general principle a DLQ entry
+  instantiates.
+
+## Retry Budgets and Backoff (P6 Backpressure is Bounded)
+
+- The retry budget is the cap on redelivery: a count, a TTL, or
+  both. After the budget, the message routes to the DLQ (P5). An
+  unbounded retry budget is the `messaging-unbounded-retry` chaos
+  anti-pattern (P5 breach): the consumer never makes progress past
+  the poison message.
+- **Exponential backoff** spaces retries: 1s, 2s, 4s, 8s, ... with
+  a jitter to avoid thundering-herd synchrony. Backoff gives the
+  downstream time to recover (P6 — the retry rate is bounded,
+  giving the downstream a chance to catch up). A retry with no
+  backoff hammers the downstream at full rate, making the failure
+  worse.
+- The retry budget × the backoff schedule is the P6 bound: the
+  consumer's retry load is bounded by design, not by luck. See
+  `domains/concurrency/patterns` Pattern 6 (Timeout on Every
+  Block) for the in-process retry/backoff analog; messaging owns
+  the broker-backed instance where the redelivery comes from the
+  broker across a network, not an in-process loop (D-062).
+
+## Poison Messages (P5, P1)
+
+- A poison message is one no retry will fix: unparseable (the
+  schema is wrong, P1), unknown type (the consumer does not handle
+  this version, P9), or a permanent failure (a not-found
+  dependency). Retrying a poison message wastes resources and
+  blocks the queue (P6 — the consumer never makes progress).
+- Poison messages route to the poison queue immediately (no
+  retry), distinct from the DLQ (which holds messages that
+  exhausted their retry budget on transient failures). The
+  distinction is the P5 discipline: a poison queue is for
+  "never going to succeed"; a DLQ is for "might have succeeded
+  but didn't within the budget."
+- A poison queue without an alert is the same defect as a DLQ
+  without an alert (P10): the operator cannot see the poison
+  accumulating. Wire both to `domains/observability/metrics`.
+
+## Cross-Link to Concurrency (P6, cross-link concurrency/patterns)
+
+- The retry/backoff discipline here is the cross-process analog
+  of `domains/concurrency/patterns` Pattern 6 (Timeout on Every
+  Block) for in-process retry. Concurrency owns the in-process
+  analog (a retry loop within one program, with a timeout per
+  attempt); messaging owns the broker-backed instance (the broker
+  redelivers across a network, the consumer applies backoff via
+  nack-with-delay). The failure model differs: in-process retry
+  fails by a thread crash or a timeout; broker-backed retry fails
+  by network partition, broker restart, or consumer crash-and-retry
+  (D-062).
+- The cross-link is one-directional outward (messaging →
+  concurrency) per D-026 extended: messaging references concurrency
+  as the in-process foundation; concurrency does not back-link to
+  messaging.
+
+## Cross-Link to Errors (P5, cross-link errors/patterns)
+
+- A poison message is an errors-as-data instance: the failure is
+  captured as a DLQ entry (with `reason`, `redeliveryCount`,
+  `deadLetteredAt`), not swallowed. See `domains/errors/patterns`
+  for the general errors-as-data discipline a DLQ entry
+  instantiates. The DLQ is the async-messaging instance of an
+  error log — observable, auditable, drainable.
+- The cross-link is one-directional outward (messaging → errors):
+  messaging references errors for the errors-as-data pattern;
+  errors does not back-link to messaging.
+
+## What Violates Delivery-Semantics Discipline
+
+| Violation | Principle |
+|-----------|-----------|
+| Unstated delivery semantic (at-least-once vs exactly-once guessed) | P4 Delivery Semantics are Explicit |
+| Non-idempotent consumer under at-least-once delivery | P3 Consumers are Idempotent |
+| Dedup store with no TTL (memory leak; grows without bound) | P6 Backpressure is Bounded |
+| Dedup key per-producer (distinct messages in the same window deduped) | P3 Consumers are Idempotent |
+| Retry with no backoff (hammers the downstream at full rate) | P6 Backpressure is Bounded |
+| Unbounded retry budget (consumer never progresses past the poison) | P5 Dead-Letter Handling is Defined |
+| DLQ with no depth alert (poison messages accumulate invisibly) | P10, `domains/observability/metrics` |
+| Poison message retried forever (no poison queue, no immediate DLQ) | P5 Dead-Letter Handling is Defined |
+| Process-not-idempotent with non-transactional mark (crash in window double-applies) | P3, P4 |
+| DLQ entry with no reason/audit metadata (uninspectable failure) | P5, `domains/errors/patterns` |
+| Ordering assumption the broker does not provide (FIFO assumed on standard queue) | P2 Ordering is a Property, Not an Assumption |
@@ -0,0 +1,307 @@
+# Messaging — First Principles
+
+## 1. The Principles
+
+### P1. Messages are Contracts
+A message has an explicit, versioned schema. Producer and consumer
+agree on shape before exchange; the schema is the boundary, not a
+guess. A schemaless message — a free-form JSON blob the consumer
+parses by hope — is a defect: the consumer breaks silently on the
+next shape change, and the producer has no contract to evolve
+against. This derives from `C1 Correctness` (the exchange must
+carry what the parties agreed to) and `C2 Clarity` (the schema
+makes the boundary obvious to both sides). This is the
+cross-process expression of the contract discipline that
+`domains/api/rest` owns for synchronous request/response: the
+message schema is to async exchange what the API contract is to
+sync exchange. It is distinct from
+`domains/concurrency/patterns` Pattern 1 (Message Passing), which
+owns the *in-process* channel primitive — here the contract spans
+separate systems and survives network failure (D-062). See
+`domains/messaging/queues.md` for the queue-flavored application
+and `domains/messaging/streams.md` for the durable-log-flavored
+application.
+
+### P2. Ordering is a Property, Not an Assumption
+Ordering guarantees — per-partition strict, global, or none — are
+explicit and documented. "It's FIFO" is a claim that must be backed
+by the broker's partitioning contract, not an assumption the
+consumer makes and the broker may not honor. A standard queue
+delivers in arrival order per receive-node but offers no global
+ordering across shards; a FIFO queue delivers strict per-message-
+group order but at a latency cost; a partitioned stream delivers
+strict per-partition order but only within a partition. Each is a
+distinct, declared property. This derives from `C1 Correctness`
+(order is a correctness property — a consumer that assumes order
+the broker does not provide is wrong) and `C2 Clarity` (the
+ordering guarantee is documented, not discovered in production).
+This is distinct from in-process ordering, which
+`domains/concurrency/patterns` Pattern 1 owns for channels within
+one program: messaging ordering survives network failure, broker
+restart, and consumer crash-and-retry — a stronger failure model
+than thread-local channels (D-062). The `messaging-shared-
+subscription` chaos anti-pattern breaches this rule: two consumers
+sharing one subscription break per-consumer ordering because the
+broker dispatches each message to an arbitrary consumer. See
+`domains/messaging/streams.md` for the partition-order contract
+and `domains/messaging/delivery-semantics.md` for the interaction
+of ordering with the three delivery semantics.
+
+### P3. Consumers are Idempotent
+Delivery is at-least-once by default across the network; a
+consumer deduplicates via idempotency keys or deterministic
+processing. "Exactly-once" is idempotency plus at-least-once, not a
+broker guarantee — Jepsen analyses of Kafka, RabbitMQ, and NATS
+establish that exactly-once claims require independent
+verification, and the durable engineering practice is to make
+consumers idempotent under redelivery. A non-idempotent consumer
+under at-least-once delivery doubles the effect on every retry; a
+non-idempotent consumer under a claimed exactly-once broker is a
+bug waiting for the broker's exactly-once invariant to break. This
+derives from `C1 Correctness`: correctness under redelivery is the
+contract, not a nice-to-have. This parallels
+`domains/edge/P5 Edge Operations are Idempotent` (the
+cross-partition device-and-cache-flavored analog) and is the
+cross-process instance of the retry-safety discipline that
+`domains/concurrency/patterns` Pattern 6 (Timeout on Every Block)
+implies for in-process retry. It is distinct from in-process
+retry because the redelivery comes from the broker across a
+network, not from an in-process loop (D-062). See
+`domains/messaging/delivery-semantics.md` for the idempotency-key
+dedup-store pattern.
+
+### P4. Delivery Semantics are Explicit
+At-least-once / at-most-once / exactly-once is a declared choice
+per channel, not an emergent behavior. The tradeoff — latency cost,
+implementation complexity, operational cost — is made consciously
+and documented. At-most-once is fire-and-forget (low latency, lossy);
+at-least-once is acked with possible duplication (the default,
+requires idempotent consumers per P3); exactly-once is at-least-once
+plus idempotency or a transactional two-phase commit (highest cost,
+narrowest fit). An unstated semantic is a defect: the consumer
+guesses, and the guess is wrong under the first failure. This
+derives from `C1 Correctness` (the chosen semantic must hold) and
+`C2 Clarity` (the tradeoff is visible to the reader and the
+operator). This is the cross-process analog of the explicit-failure-
+mode discipline that `domains/errors/patterns` owns for synchronous
+code — messaging makes the delivery-mode choice as explicit as an
+error-handling choice. See `domains/messaging/queues.md` for the
+three-semantics comparison table and `domains/messaging/delivery-
+semantics.md` for the correctness properties of each.
+
+### P5. Dead-Letter Handling is Defined
+Poison messages — unparseable, repeatedly failing, or exhausting
+the retry budget — are routed to a dead-letter queue, not retried
+forever or silently dropped. The DLQ is observable and drainable: an
+operator can inspect it, replay from it, or discard with audit. An
+unbounded retry loop is a livelock: the consumer never makes
+progress past the poison message. A silent drop is a correctness
+defect: the message vanished with no record. This derives from `C1
+Correctness` (poison messages must not livelock the consumer or
+silently disappear) and `C5 Reversibility` (the DLQ is the
+reversibility mechanism — a dead-lettered message can be reprocessed
+after the bug is fixed). This is the cross-process analog of the
+bounded-error discipline that `domains/errors/patterns` owns for
+synchronous code: a poison message is an error-as-data instance
+that must be observable and recoverable, not swallowed. It is
+distinct from in-process error handling because the failure spans a
+network and a consumer restart (D-062). See
+`domains/messaging/delivery-semantics.md` for the dead-letter
+strategy comparison table and the DLQ routing rule pattern.
+
+### P6. Backpressure is Bounded
+A slow consumer cannot unbounded-buffer the broker or the
+producer. Backpressure is explicit: consumer lag is visible,
+max-unacked is bounded, the retry budget is capped. A consumer
+that falls behind without a visible signal is a silent backlog —
+the operator cannot fix what they cannot see, and the broker's
+memory grows without bound until it fails. This derives from `C1
+Correctness` (a backlog that grows until OOM is a correctness
+failure) and `C8 Economy` (the broker's memory is bounded by
+design, not by luck). This is distinct from
+`domains/concurrency/P9 Bounded Queues` and
+`domains/concurrency/patterns` Pattern 5 (Bounded Queue with
+Backpressure), which own the *in-process* analog: concurrency's
+bounded queue fails by OOM or thread crash; messaging's bounded
+backpressure fails by network partition, broker restart, or
+consumer crash-and-retry (D-062). The Reactive Streams
+specification (`request(n)`, `onNext` bounded) is the in-process
+instance; messaging's broker-backed backpressure is the
+cross-process instance above it. See `domains/messaging/queues.md`
+for prefetch and max-unacked and `domains/observability/metrics`
+for consumer-lag as an alert.
+
+### P7. Partitioning is Intentional
+The partition key determines ordering, parallelism, and hotspots.
+Key choice is a design decision with documented rationale, not a
+default. A key that hashes unevenly creates a hot partition that
+limits throughput; a key that does not match the ordering need
+breaks per-key semantics; a key that is too coarse (one partition
+for the whole topic) serializes all the traffic. The partition
+count is a capacity bound: too few partitions cap parallelism, too
+many partition overhead the broker. This derives from `C4 Locality`
+(ordering and parallelism are co-located with the partition) and
+`C6 Composability` (the partition is the unit of parallelism and
+scaling — consumer groups compose from per-partition workers).
+This is the cross-process analog of the locality discipline that
+`domains/performance/` owns for generic data-near-compute
+optimization: performance's locality is algorithmic (data near
+compute); messaging's locality is partitional (order and
+parallelism near the partition). See `domains/messaging/streams.md`
+for the partitioned-log model and consumer-group rebalance
+strategies.
+
+### P8. Replay and Retention are Configured
+Retention windows and replay-from-offset are explicit. A message
+is not ephemeral by default; the broker is a durable log, not a
+pipe. A topic with no retention is a fire-and-forget stream — a
+consumer that falls behind loses data permanently; a topic with
+infinite retention is an unbounded log — the broker grows until
+disk exhaustion. Both are defects: the retention window is a
+declared bound, and replay-from-offset is the mechanism that makes
+the log durable (re-consumable) rather than ephemeral. This derives
+from `C5 Reversibility` (a retained message is reversible — it can
+be re-consumed; an ephemeral message is not) and `C7 Observability`
+(the durable log is itself an observable record of what happened —
+the offset is the position from which to replay). This is the
+foundation for `domains/messaging/streams.md` and the rule that
+distinguishes a stream from a queue (a queue deletes on ack; a
+stream retains for replay). See `domains/messaging/pubsub.md` for
+the pub/sub-vs-stream durability boundary.
+
+### P9. Schemas Evolve Compatibly
+Schema changes are backward- and forward-compatible by
+construction. Breaking changes are versioned migrations, not
+silent shape edits. A producer that ships a new field the old
+consumer ignores is backward-compatible; a consumer that handles a
+missing field the new producer omits is forward-compatible. A
+silent schema change — the producer renames a field and the
+consumer parses `undefined` — is a P1 violation (the contract was
+broken) compounded here as an evolution defect. This derives from
+`C5 Reversibility` (a schema change is reversible by versioning —
+the old shape is still readable) and `C6 Composability` (producers
+and consumers of different versions compose because the schema
+evolves compatibly). This parallels `domains/data/migrations`
+(schema migration for databases) and `domains/api/versioning`
+(API contract evolution): messaging's schema evolution is the
+async instance of the same compatibility discipline. See
+`domains/messaging/streams.md` for the stream-schema-evolution
+angle.
+
+### P10. Messaging is Observable
+Consumer lag, DLQ depth, throughput, and consumer-group health are
+first-class signals. Silent backlog is a bug, not a feature: a
+consumer that falls behind with no lag metric is invisible until
+the downstream effect surfaces — by which time the backlog may be
+hours or days. A DLQ that grows without an alert is a silent
+correctness defect: poison messages are accumulating and no one
+knows. This derives from `C7 Observability` (the broker's behavior
+is visible to the operator) and `C1 Correctness` (backlog
+detection is a correctness bound — unbounded lag is a failure).
+This is distinct from `domains/observability/metrics`, which owns
+*generic* structured metrics; messaging owns the *broker-specific*
+signals — lag, DLQ depth, partition imbalance, consumer-group
+rebalance events. See `domains/observability/metrics` for the
+generic SLI/SLO discipline and `domains/observability/tracing` for
+cross-partition traces.
+
+## 2. Core Principle Trace
+
+Each messaging P-rule derives from one or more core C-rules
+(C1–C8). The matrix extension lands in P4 of the v0.4 plan; the
+traces below are authoritative. Messaging is a broad-derivation
+domain touching 7 of 8 core principles (C1, C2, C4, C5, C6, C7,
+C8); C3 (Simplicity) is not a primary derivation — messaging is
+inherently a tradeoff domain where simplicity yields to the
+correctness of delivery guarantees (a simpler-than-necessary
+delivery model does not handle the failure cases, per C3's
+"simpler than necessary is also a violation").
+
+| P-rule | Core | Why |
+|--------|------|-----|
+| P1 Messages are Contracts | C1, C2 | Correctness of the exchange; clarity of the schema boundary |
+| P2 Ordering is a Property, Not an Assumption | C1, C2 | Correctness of order; clarity of the guarantee |
+| P3 Consumers are Idempotent | C1 | Correctness under redelivery |
+| P4 Delivery Semantics are Explicit | C1, C2 | Correctness of the chosen semantic; clarity of the tradeoff |
+| P5 Dead-Letter Handling is Defined | C1, C5 | Correctness of poison-message routing; reversibility of reprocessing |
+| P6 Backpressure is Bounded | C1, C8 | Correctness of bounded backlog; economy of broker memory |
+| P7 Partitioning is Intentional | C4, C6 | Locality of order; composability of parallelism |
+| P8 Replay and Retention are Configured | C5, C7 | Reversibility of replay; observability of the durable log |
+| P9 Schemas Evolve Compatibly | C5, C6 | Reversibility of schema changes; composability of versions |
+| P10 Messaging is Observable | C7, C1 | Observability of lag/DLQ; correctness of backlog detection |
+
+## 3. What Violates These Principles
+
+| Violation | Principle Breached |
+|-----------|-------------------|
+| Schemaless message (no versioned contract; consumer parses by guess) | P1 Messages are Contracts |
+| "It's FIFO" with no documented partition contract | P2 Ordering is a Property, Not an Assumption |
+| Non-idempotent consumer under at-least-once delivery | P3 Consumers are Idempotent |
+| Unstated delivery semantic (at-least-once vs exactly-once guessed) | P4 Delivery Semantics are Explicit |
+| No dead-letter queue (poison message retried forever or silently dropped) | P5 Dead-Letter Handling is Defined |
+| Unbounded retry budget (no cap; slow consumer stalls the partition) | P6 Backpressure is Bounded |
+| Default partition key (no rationale; hotspot or wrong-order) | P7 Partitioning is Intentional |
+| Ephemeral broker (no retention; no replay) | P8 Replay and Retention are Configured |
+| Silent schema change (producer breaks consumers with no version bump) | P9 Schemas Evolve Compatibly |
+| Silent backlog (no lag metric; consumer falls behind invisibly) | P10 Messaging is Observable |
+| Shared subscription (two consumers share one subscription; per-consumer ordering breaks) | P2 Ordering is a Property, Not an Assumption (P3 compounding) |
+| Blocking consumer (slow downstream call with no timeout; broker redelivers to the stuck consumer) | P6 Backpressure is Bounded |
+
+## 4. Relationship to Other Domains
+
+Messaging systems are the engineering discipline of
+**cross-process, cross-system asynchronous communication via
+brokers**. Producer and consumer are separate systems; the broker
+is the intermediary that brokers delivery, ordering, retention,
+and failure semantics. The distinguishing constraints are a
+cross-process failure model (network, not crash), explicit
+delivery semantics, decoupled producer/consumer lifecycle, and
+replay-and-retention as a durable-log property. Messaging overlaps
+`domains/concurrency/` by *subject* (messages, queues,
+backpressure) but not by *failure model*: per D-062, messaging
+owns the cross-process/network-failure-model angle; concurrency
+owns the in-process/crash-failure-model angle. The discriminator
+is the failure model: concurrency's queue fails by OOM or thread
+crash; messaging's queue fails by network partition, broker
+restart, or consumer crash-and-retry. Messaging extends
+concurrency's bounded-queue/backpressure model to the network-
+partition regime. Cross-links are one-directional outward (per
+D-026 extended); no back-link edits to v0.1/v0.2/v0.3 content.
+
+- `domains/concurrency/patterns` ← P6 (the broker-backed bounded
+  queue is the cross-process analog of the in-process bounded
+  buffer — concurrency Pattern 5 owns in-process; messaging owns
+  the network-failure-model instance above it, per D-062)
+- `domains/concurrency/patterns` ← P3 (idempotent retry is the
+  cross-process analog of in-process retry-safety — the failure
+  model differs: broker redelivery across a network vs in-process
+  loop)
+- `domains/observability/metrics` ← P10 (consumer lag and DLQ
+  depth as alerts; observability owns the generic SLI/SLO
+  discipline, messaging owns the broker-specific signals)
+- `domains/observability/tracing` ← P10 (cross-partition traces
+  for stream processing; observability owns the generic tracing
+  discipline, messaging owns the cross-partition propagation)
+- `domains/data/schema-design` ← P1, P9 (message schema design
+  and evolution; data owns the generic schema discipline,
+  messaging owns the cross-process message-shape instance)
+- `domains/errors/patterns` ← P5 (errors as data for message
+  failures; a poison message is an error-as-data instance that must
+  be observable and recoverable, not swallowed)
+- `domains/edge/iot` ← P4 (the edge↔messaging cross-link
+  resolves bidirectionally here: edge/iot.md links outward to
+  messaging/queues for MQTT QoS parallels to delivery semantics;
+  this first-principles doc acknowledges the back-link — the
+  edge/iot.md → messaging/queues link from P1 now resolves because
+  messaging/queues.md exists, completing the bidirectionality per
+  IDEATE-40)
+
+> Note: the edge/iot.md → messaging/queues cross-link (MQTT QoS
+> parallels for delivery semantics) was authored in P1 with a
+> dangling reference; this P2 authorship of messaging/queues.md
+> resolves it. The bidirectionality is verified in P5
+> (ATELIER-114 per IDEATE-40). The cross-link is one-directional
+> outward from edge/iot.md; this first-principles doc
+> acknowledges the resolution without editing edge/iot.md (per
+> D-026 extended — no back-link edits to v0.1/v0.2/v0.3 or to
+> P1-authored edge content).
@@ -0,0 +1,269 @@
+# Pub/Sub — Derived Rules
+
+> Derives from `domains/messaging/first-principles.md`. Applies P1
+> (Messages are Contracts), P2 (Ordering is a Property, Not an
+> Assumption), P3 (Consumers are Idempotent), and P4 (Delivery
+> Semantics are Explicit) primarily, with P7 (partitioning), P10
+> (per-subscription lag). The `messaging-shared-subscription` chaos
+> anti-pattern lives here (pre-specified in P4 ATELIER-110).
+> Cross-links `domains/messaging/streams` for the pub/sub-vs-stream
+> durability boundary and `domains/observability/metrics` for
+> per-subscription lag.
+
+## What Pub/Sub Is (P1 Messages are Contracts)
+
+- Pub/sub is the fan-out primitive: a producer publishes a message
+  to a topic; N independent subscriptions each receive a copy. The
+  message has an explicit, versioned schema (P1): the topic's
+  schema is the contract every subscription agrees to before
+  subscribing. A schemaless topic is a defect — every subscriber
+  breaks silently on the next shape change.
+- The boundary with queues is the fan-out ratio. A queue is
+  point-to-point (one producer, one consumer); pub/sub is
+  one-to-many (one producer, N consumers, each with its own
+  subscription). The boundary with streams is the durability model
+  — see the cross-link below. Pub/sub is an async concern because
+  producer and consumers are separate systems and the failure model
+  is network, not crash (D-062).
+- See `domains/messaging/queues.md` for the point-to-point variant
+  and `domains/messaging/streams.md` for the durable-log variant.
+
+## Topic / Subscription Model (P1, P3, P4)
+
+- A **topic** is the named stream of messages. A **subscription**
+  is a durable cursor over the topic: each subscription receives
+  every message published after it was created (subject to
+  retention and filtering). The subscription is independent — its
+  ack, redelivery, and DLQ are per-subscription, not shared.
+- Each subscription is a consumer under at-least-once by default
+  (P4): the broker redelivers until the subscription acks, and the
+  subscriber must be idempotent (P3). A subscription with no
+  idempotency dedup duplicates every redelivered message.
+- The topic's schema evolves compatibly (P9 — Schemas Evolve
+  Compatibly): a new field the old subscriber ignores is
+  backward-compatible; a renamed field the old subscriber parses
+  as `undefined` is a P1 violation.
+
+```python
+# Publish + two independent subscriptions (P1 contract, P3
+# idempotency, P4 at-least-once per subscription). Each
+# subscription is an independent durable cursor; acking one does
+# not affect the other.
+
+import json, uuid
+
+# --- Publisher ---
+def publish(topic, event, broker):
+    # P1: versioned schema on the topic. All subscribers must
+    # understand this schema (or a compatible superset — P9).
+    message = {
+        "schema": "user.signed-up.v1",
+        "id": str(uuid.uuid4()),
+        "idempotencyKey": f"user:{event['userId']}:signup",
+        "payload": event,
+    }
+    broker.publish(topic=topic, body=json.dumps(message))
+
+# --- Subscription A: welcome-email service ---
+def subscribe_welcome(broker, dedup_store, send_email):
+    sub = broker.subscribe(topic="users", subscription="welcome-email")
+    for message in sub.receive():
+        # P3: idempotent per subscription. A redelivered message is
+        # a no-op for THIS subscription, not for the others.
+        if dedup_store.seen(("welcome", message["idempotencyKey"])):
+            sub.ack(message)
+            continue
+        send_email(message["payload"]["email"], "Welcome!")
+        dedup_store.mark(("welcome", message["idempotencyKey"]))
+        sub.ack(message)
+
+# --- Subscription B: analytics-ingest service ---
+def subscribe_analytics(broker, dedup_store, ingest):
+    # Independent subscription: its own cursor, its own dedup,
+    # its own ack. Welcome-email acking does NOT advance this.
+    sub = broker.subscribe(topic="users", subscription="analytics")
+    for message in sub.receive():
+        if dedup_store.seen(("analytics", message["idempotencyKey"])):
+            sub.ack(message)
+            continue
+        ingest(message["payload"])
+        dedup_store.mark(("analytics", message["idempotencyKey"]))
+        sub.ack(message)
+```
+
+- The dedup key is scoped per subscription: `(subscription,
+  idempotencyKey)`. A redelivery to subscription A that was already
+  processed by A is a no-op for A; the same message delivered to
+  subscription B is processed by B independently. Scoping the dedup
+  key by subscription prevents one subscription's dedup from
+  masking another's redelivery.
+
+## Fan-Out Semantics (P4, P7)
+
+- Fan-out means every subscription receives every published message
+  (subject to filtering — see below). The broker duplicates the
+  message per subscription; each subscription's delivery is
+  independent. The fan-out ratio is the number of subscriptions; the
+  broker's cost scales with fan-out × message size.
+- Partitioning (P7) applies to topics that are partitioned for
+  throughput: a partitioned topic delivers per-partition order, and
+  each subscription receives from every partition. A subscription
+  that consumes partitions in parallel must handle per-partition
+  ordering and cross-partition non-ordering (P2 — document the
+  property, do not assume global order).
+- The delivery semantic is per-subscription (P4): subscription A
+  may be at-least-once, subscription B may be at-most-once (for a
+  loss-tolerant analytics feed). The choice is per subscription,
+  declared, not emergent.
+
+## Shared vs Independent Subscriptions (P2, P3 — the chaos anti-pattern)
+
+- An **independent subscription** is one durable cursor per
+  consumer group: each subscription receives every message in
+  topic order (per partition, P2) and acks independently. This is
+  the correct default: per-consumer ordering and per-consumer
+  idempotency hold.
+- A **shared subscription** is one subscription shared by multiple
+  consumers: the broker dispatches each message to an arbitrary
+  consumer in the shared group. This breaks per-consumer ordering
+  (P2 — consumer A sees message 3 before consumer B sees message
+  1) and complicates idempotency (P3 — the dedup state must be
+  shared across consumers, not per-consumer). This is the
+  `messaging-shared-subscription` chaos anti-pattern
+  (pre-specified in P4 ATELIER-110): the primary breach is P2
+  (ordering); P3 (idempotency) is the compounding consequence.
+- A shared subscription is correct ONLY when the consumers are
+  stateless, the per-message processing is order-independent, and
+  the dedup store is shared (a shared Redis, a shared DB). A shared
+  subscription for order-dependent or per-consumer-stateful
+  processing is the chaos anti-pattern: the broker's arbitrary
+  dispatch breaks the order the consumer assumes.
+
+```python
+# The messaging-shared-subscription chaos anti-pattern (P2
+# ordering breach, P3 idempotency compounding). Two consumers
+# share one subscription; the broker dispatches each message to
+# an arbitrary consumer. Per-consumer ordering breaks; dedup
+# must be shared (and often is not).
+
+# BAD — shared subscription, per-consumer dedup (chaos):
+def shared_subscription_bad(broker, send_email):
+    # Both consumers call subscribe with the SAME subscription
+    # name. The broker round-robins; consumer A gets msg 1, msg 3;
+    # consumer B gets msg 2, msg 4. Per-consumer order is broken.
+    # If each consumer has its OWN dedup store, a redelivery to
+    # the OTHER consumer re-processes (P3 breach).
+    sub = broker.subscribe(topic="users", subscription="shared")
+    for message in sub.receive():
+        # Per-consumer dedup — WRONG. A redelivered message may
+        # land on the other consumer, which has not seen it.
+        if local_dedup.seen(message["idempotencyKey"]):  # per-consumer
+            sub.ack(message); continue
+        send_email(message["payload"]["email"], "Welcome!")
+        local_dedup.mark(message["idempotencyKey"])
+        sub.ack(message)
+
+# CORRECT — independent subscriptions (per-consumer ordering,
+# per-subscription dedup):
+def independent_subscriptions_good(broker, send_email):
+    sub = broker.subscribe(topic="users", subscription="welcome-email")
+    for message in sub.receive():
+        if dedup_store.seen(("welcome", message["idempotencyKey"])):
+            sub.ack(message); continue
+        send_email(message["payload"]["email"], "Welcome!")
+        dedup_store.mark(("welcome", message["idempotencyKey"]))
+        sub.ack(message)
+```
+
+- If a shared subscription is genuinely required (stateless,
+  order-independent, shared dedup), document the choice and the
+  shared-dedup requirement (P2 — the ordering property is "none
+  across consumers"; P3 — the dedup is shared). The default is
+  independent subscriptions; shared is an opt-in for the narrow case.
+
+## Filtering (P4, C8 Economy)
+
+- **Subscription filtering** lets a subscription receive only
+  messages matching a filter (e.g., `event.type == "order"`).
+  Filtering at the broker saves bandwidth (C8 — the subscriber
+  does not receive and discard) and reduces subscriber load.
+- **Server-side filtering** (broker evaluates the filter before
+  delivery) is more efficient than **client-side filtering**
+  (subscriber receives and discards). Server-side filtering is the
+  default where the broker supports it (GCP Pub/Sub, SNS filtering,
+  NATS subject filtering); client-side is the fallback.
+- A filter that is too broad wastes bandwidth; a filter that is
+  too narrow drops messages the subscriber needed. The filter is
+  a P1 (contract) and P4 (semantic) decision: the subscription's
+  filter is part of its declared contract.
+
+## Ordering Across Subscriptions (P2)
+
+- A topic with per-partition ordering delivers per-partition order
+  to each subscription. Across subscriptions, there is no ordering
+  guarantee: subscription A may ack message 3 while subscription B
+  is still on message 1. This is correct and expected — each
+  subscription is independent.
+- Within a subscription, ordering holds per partition (P2 — the
+  documented property). A subscription that processes partitions
+  in parallel must not assume cross-partition order. A subscription
+  that needs global order must use a single partition (sacrificing
+  parallelism, P7) or an external sequencing mechanism.
+- The `messaging-shared-subscription` anti-pattern breaks even
+  per-partition order within a subscription: the broker's arbitrary
+  dispatch to consumers in the shared group breaks the per-
+  partition sequence each consumer sees.
+
+## Pub/Sub vs Stream — The Durability Boundary (cross-link messaging/streams)
+
+- Pub/sub and streams are both fan-out or one-to-many primitives,
+  but their durability model differs. Pub/sub is a
+  **push-to-subscription** model: each subscription is a cursor,
+  retention is short (the subscription's unacked window), and
+  replay is limited to the unacked messages. A subscription that
+  falls behind beyond the retention window loses messages
+  permanently.
+- A stream is a **durable-log** model: messages are retained by
+  the log for a configured window (P8 — Replay and Retention are
+  Configured), and any consumer group can replay from any offset
+  within the window. A stream consumer that falls behind can
+  catch up by replaying; a pub/sub subscription that falls behind
+  beyond retention cannot.
+- The choice is the durability requirement: if the consumer must
+  be able to replay (reprocessing, backfill, new consumer starting
+  from the beginning), use a stream. If the consumer only needs
+  the live feed (and can tolerate loss on a long fall-behind),
+  pub/sub is lighter. See `domains/messaging/streams.md` for the
+  durable-log model, offsets, and consumer groups.
+
+## Observability — Per-Subscription Lag (P10)
+
+- Per-subscription lag (messages published minus messages acked
+  for each subscription, or the age of the oldest unacked message
+  per subscription) is the primary pub/sub health signal. Each
+  subscription has its own lag — a fast subscription and a slow
+  subscription on the same topic are independent signals.
+- A subscription whose lag grows beyond the retention window is a
+  silent data-loss risk: the broker will drop the oldest messages,
+  and the subscription will never see them. Alert on lag relative
+  to retention — lag approaching retention is the loss threshold.
+- Wire per-subscription lag to `domains/observability/metrics` as
+  an SLI per subscription. A topic with N subscriptions has N lag
+  metrics; a single aggregate hides the slow one. See
+  `domains/observability/metrics` for the generic SLI/SLO
+  discipline.
+
+## What Violates Pub/Sub Discipline
+
+| Violation | Principle |
+|-----------|-----------|
+| Shared subscription for order-dependent processing (broker dispatch breaks per-consumer order) | P2 Ordering is a Property, Not an Assumption |
+| Shared subscription with per-consumer dedup (redelivery to the other consumer re-processes) | P3 Consumers are Idempotent |
+| Schemaless topic (no versioned contract; subscribers parse by guess) | P1 Messages are Contracts |
+| Subscription with no idempotency dedup (redelivered message duplicates the effect) | P3 Consumers are Idempotent |
+| Subscription whose lag approaches retention (silent data loss) | P10, `domains/observability/metrics` |
+| Unstated delivery semantic per subscription (at-least-once vs at-most-once guessed) | P4 Delivery Semantics are Explicit |
+| Filter that is too narrow (drops messages the subscriber needed) | P1, P4 |
+| Partitioned topic with no documented per-partition ordering contract | P2 Ordering is a Property, Not an Assumption |
+| No per-subscription lag metric (slow subscription invisible) | P10 Messaging is Observable |
+| Cross-partition order assumption within a subscription (no global order guarantee) | P2, P7 |
@@ -0,0 +1,318 @@
+# Queues — Derived Rules
+
+> Derives from `domains/messaging/first-principles.md`. Applies P1
+> (Messages are Contracts), P3 (Consumers are Idempotent), P4
+> (Delivery Semantics are Explicit), P5 (Dead-Letter Handling is
+> Defined), and P6 (Backpressure is Bounded) primarily, with P2
+> (ordering), P7 (partitioning), and P10 (observable lag). For the
+> at-least-once / at-most-once / exactly-once decision, see the
+> comparison table below. Cross-links `domains/concurrency/patterns`
+> for the in-process bounded-queue analog and
+> `domains/observability/metrics` for consumer lag.
+
+## What a Queue Is (P1 Messages are Contracts)
+
+- A queue is a point-to-point async delivery primitive. A producer
+  enqueues a message; exactly one consumer dequeues and processes
+  it. The message has an explicit, versioned schema (P1): the
+  producer and consumer agree on shape before exchange, and the
+  schema is the boundary — a schemaless message is a defect (the
+  consumer breaks silently on the next shape change).
+- The boundary is per D-062: messaging owns the cross-process /
+  network-failure-model angle; concurrency owns the in-process
+  analog. A queue is a messaging concern because producer and
+  consumer are separate systems, the broker is the intermediary,
+  and the failure model is network (the message can be lost,
+  duplicated, reordered, or delayed by the broker or the network,
+  not by a thread crash). The in-process bounded buffer
+  (`domains/concurrency/patterns` Pattern 5) is the analog below
+  this boundary — it fails by OOM; a broker-backed queue fails by
+  partition, broker restart, or consumer crash-and-retry.
+- See `domains/messaging/pubsub.md` for the fan-out (one-to-many)
+  variant and `domains/messaging/streams.md` for the durable-log
+  (replay-from-offset) variant. A queue deletes on ack; a stream
+  retains for replay — the durability boundary is the
+  distinguishing trait.
+
+## Producer / Consumer Model (P1, P3)
+
+- The producer enqueues a message with an idempotency key (P3).
+  The consumer dequeues, processes, and acks. If the consumer
+  crashes before acking, the broker redelivers; the idempotency
+  key makes the redelivery safe (the consumer dedups, not the
+  broker).
+- The idempotency key is per-message, not per-producer or
+  per-session. A consumer that dedups by producer alone will drop
+  distinct messages issued in the same window. Use a UUID per
+  message, or a deterministic key derived from the message content
+  (e.g., `(entity, operation, version)`).
+
+```python
+# Producer/consumer pair with idempotency key (P1 contract, P3
+# idempotency). The producer tags each message with a versioned
+# schema and a unique idempotency key; the consumer dedups by the
+# key so a redelivered message is processed once (P3).
+
+# --- Producer ---
+import json, uuid
+
+def enqueue(order, broker):
+    # P1: versioned schema. The message carries its schema version
+    # so the consumer can route by shape (P9 evolution discipline).
+    message = {
+        "schema": "order.created.v1",
+        "id": str(uuid.uuid4()),
+        "idempotencyKey": f"order:{order['id']}:{order['version']}",
+        "payload": order,
+    }
+    broker.send(queue="orders", body=json.dumps(message))
+    # At-least-once by default (P4): the broker acks the send; the
+    # consumer may see this message more than once under retry.
+
+# --- Consumer ---
+def consume(broker, dedup_store, process_order):
+    for message in broker.receive(queue="orders"):
+        # P3: idempotent consumer. Dedup by idempotency key before
+        # processing; a redelivered message is a no-op, not a
+        # double-apply.
+        if dedup_store.seen(message["idempotencyKey"]):
+            broker.ack(message)        # already processed; skip
+            continue
+        try:
+            process_order(message["payload"])
+            dedup_store.mark(message["idempotencyKey"])
+            broker.ack(message)        # success; broker drops it
+        except Exception:
+            broker.nack(message)       # redeliver (at-least-once, P4)
+```
+
+- The dedup store is bounded (P6 — Backpressure is Bounded): a
+  dedup store that grows without bound is a memory leak. Use a TTL
+  window longer than the broker's max-redelivery window, or a
+  bounded LRU. See `domains/messaging/delivery-semantics.md` for
+  the full idempotency-key dedup-store pattern.
+
+## Ack / Nack (P4 Delivery Semantics are Explicit)
+
+- **Ack** tells the broker the message was processed; the broker
+  drops it. **Nack** (negative ack) tells the broker the
+  processing failed; the broker redelivers (at-least-once) or
+  routes to a DLQ (after the retry budget — P5).
+- A consumer that neither acks nor nacks within the visibility
+  timeout causes the broker to redeliver (the broker assumes the
+  consumer died). This is the at-least-once default: the broker
+  prefers duplication to loss.
+- The semantic is explicit (P4): at-least-once is the default; the
+  consumer must be idempotent (P3). At-most-once is fire-and-forget
+  (no ack; the broker drops on send) — lossy but lowest latency.
+  Exactly-once is at-least-once plus idempotency, or a
+  transactional two-phase commit — see the comparison table below.
+
+## Visibility Timeouts and Redelivery (P4, P5)
+
+- The visibility timeout is the window the broker hides a message
+  after delivery, waiting for the ack. If the consumer does not
+  ack within the window, the broker makes the message visible
+  again and redelivers it (to the same consumer or another). This
+  is the at-least-once mechanism: the broker assumes a
+  no-ack-in-time consumer is dead.
+- The timeout must be longer than the processing time, or the
+  broker redelivers a message the consumer is still processing —
+  causing duplicate processing (which P3 idempotency makes safe,
+  but which wastes resources). A timeout shorter than processing
+  time is a P6 (backpressure) smell: the consumer is too slow for
+  the configured timeout.
+- Redelivery has a budget (P5): after N redeliveries or a TTL, the
+  message routes to the DLQ. An unbounded retry budget is the
+  `messaging-unbounded-retry` chaos anti-pattern: the consumer
+  never makes progress past the poison message.
+
+## FIFO vs Standard Queues (P2 Ordering is a Property, Not an Assumption)
+
+- A **standard queue** delivers in arrival order per receive-node
+  but offers no global ordering across shards, no per-message-
+  group ordering, and may redeliver out of order under retry. It
+  is the high-throughput default; ordering is *not* guaranteed
+  (P2: the ordering property is "none" — explicitly documented).
+- A **FIFO queue** delivers strict per-message-group order: all
+  messages with the same group ID are delivered to one consumer
+  in send order. The cost is throughput (FIFO queues cap at lower
+  TPS) and latency (the broker must sequence per group). The
+  ordering property is "per-group strict" — explicitly documented
+  (P2).
+- The choice is a P2 decision (which ordering guarantee) and a C8
+  decision (throughput cost). A consumer that assumes FIFO on a
+  standard queue is a P2 violation: the broker does not provide
+  the guarantee the consumer assumes. Document the property; do
+  not assume it.
+
+## Prefetch and Concurrency (P6 Backpressure is Bounded)
+
+- **Prefetch** (or max-unacked) bounds how many messages the
+  broker delivers to one consumer without an ack. A prefetch of 1
+  is strict stop-and-wait (lowest throughput, tightest backpressure);
+  a prefetch of N allows the consumer to process N in flight
+  (higher throughput, more memory). An unbounded prefetch is a P6
+  violation: the broker floods the consumer's memory.
+- **Consumer concurrency** is the number of parallel workers
+  processing from the queue. More workers increase throughput up to
+  the downstream's limit; beyond that, the workers stall the
+  downstream (P6 — the backpressure propagates to the
+  downstream, not the broker).
+- The prefetch × concurrency product is the in-flight cap. Declare
+  it (P6): an undeclared cap is a defect — the consumer either
+  underutilizes the broker (prefetch too low) or OOMs under load
+  (prefetch too high). This is the cross-process analog of
+  `domains/concurrency/patterns` Pattern 5 (bounded queue with
+  backpressure): concurrency owns the in-process analog;
+  messaging owns the broker-backed instance.
+
+## Long Polling (P6, C8 Economy)
+
+- Long polling (or `ReceiveMessage` with a wait-time-seconds)
+  holds the receive request open until a message arrives or the
+  wait expires. This reduces empty-receive round trips (C8
+  economy of the constrained link) and reduces latency-to-first-
+  message (the message is delivered when it arrives, not on the
+  next poll cycle).
+- Long polling is the default for low-throughput queues: short
+  polling burns CPU on empty receives; long polling waits for
+  work. For high-throughput queues, the broker is usually full
+  enough that long polling adds no latency; for low-throughput
+  queues, long polling is the difference between 20ms and 20s
+  latency-to-first-message.
+
+## Redelivery + DLQ Flow (P5 Dead-Letter Handling is Defined)
+
+- A poison message (unparseable, repeatedly failing, or exhausting
+  the retry budget) routes to the dead-letter queue. The DLQ is
+  observable (P10 — DLQ depth is an alert) and drainable (an
+  operator can inspect, replay, or discard with audit).
+- The retry budget is bounded (P6): N redeliveries, or a TTL with
+  exponential backoff. After the budget is exhausted, the message
+  is moved to the DLQ, not retried forever. An unbounded retry is
+  the `messaging-unbounded-retry` chaos anti-pattern (P5 breach).
+- The DLQ routing rule is a redelivery-count or TTL threshold
+  plus a target queue. See `domains/messaging/delivery-semantics.md`
+  for the dead-letter strategy comparison table.
+
+```python
+# Redelivery + DLQ flow (P5 dead-letter handling, P6 bounded
+# retry budget). The consumer tracks redelivery count; after the
+# budget, the message routes to the DLQ. The DLQ is observable
+# (P10 — depth is an alert) and drainable.
+
+MAX_REDELIVERIES = 5
+DLQ = "orders-dlq"
+
+def consume_with_dlq(broker, dedup_store, process_order):
+    for message in broker.receive(queue="orders"):
+        # P3 idempotency: a redelivered, already-processed message
+        # is acked and skipped (not re-processed, not DLQ'd).
+        if dedup_store.seen(message["idempotencyKey"]):
+            broker.ack(message)
+            continue
+        try:
+            process_order(message["payload"])
+            dedup_store.mark(message["idempotencyKey"])
+            broker.ack(message)
+        except Exception as exc:
+            # P5: bounded retry budget. After MAX_REDELIVERIES,
+            # route to DLQ — do NOT retry forever.
+            count = message.get("redeliveryCount", 0) + 1
+            if count >= MAX_REDELIVERIES:
+                broker.send(DLQ, body=json.dumps({
+                    "original": message,
+                    "reason": str(exc),
+                    "deadLetteredAt": now_iso(),
+                    "redeliveryCount": count,
+                }))
+                broker.ack(message)  # remove from the origin queue
+                # P10: the DLQ depth must alert. A DLQ that grows
+                # with no alert is a silent correctness defect.
+            else:
+                # Nack with backoff: the broker redelivers after a
+                # delay. The backoff caps the retry rate (P6).
+                broker.nack(message, delay=exponential_backoff(count))
+```
+
+- The `deadLetteredAt` and `reason` fields make the DLQ entry
+  observable and auditable: an operator inspecting the DLQ sees
+  why each message was dead-lettered and when. See
+  `domains/errors/patterns` for the errors-as-data discipline the
+  DLQ entry follows.
+
+## Delivery Semantics Comparison (D-069)
+
+| Semantic | Guarantee | Latency Cost | Implementation Cost | When It Fits |
+|----------|-----------|-------------|---------------------|--------------|
+| **At-most-once** | A message is delivered 0 or 1 times; loss is possible, duplication is not | Lowest (no ack; fire-and-forget) | Lowest (no ack, no dedup) | Telemetry where a dropped sample is acceptable; high-throughput metrics; MQTT QoS 0; logs where a lost line is tolerable. Never for billing, orders, or any side-effecting operation. |
+| **At-least-once** | A message is delivered 1 or more times; duplication is possible, loss is not | Low (one ack round-trip) | Medium (consumer must be idempotent — P3; dedup store required) | The default for side-effecting operations: orders, payments, commands. The consumer dedups via idempotency keys (P3); the broker guarantees delivery. Fits the vast majority of broker-backed queues (SQS standard, RabbitMQ ack, MQTT QoS 1). |
+| **Exactly-once** | A message is delivered exactly 1 time; no loss, no duplication | Highest (two-phase commit or transactional producer+consumer) | Highest (requires transactions, a transactional producer, and a transactional consumer — or at-least-once plus idempotency, which collapses to at-least-once with dedup) | Rare. Kafka transactions (consume-process-produce in one transaction); MQTT QoS 2 (four-step handshake). In practice, "exactly-once" is usually at-least-once plus idempotency (P3) — the broker does not guarantee it; the consumer enforces it. Jepsen analyses verify broker claims. |
+
+- The default for side-effecting operations is **at-least-once with
+  idempotent consumers** (P3). At-most-once is for loss-tolerant
+  telemetry. Exactly-once is reserved for the narrow case where
+  the consume-process-produce loop must be transactional (Kafka
+  transactions) — and even then, the consumer should be idempotent
+  as defense-in-depth.
+- The latency cost column is the C8 tradeoff: at-most-once is
+  cheapest, exactly-once is most expensive. The implementation
+  cost column is the C1/C3 tradeoff: at-most-once is simplest,
+  exactly-once is most complex (and most fragile — a transactional
+  consumer that partially fails is a bug source). The "when it
+  fits" column is the P4 decision: declare the semantic per
+  channel, do not let it emerge.
+- This is the decision matrix required by D-069 for queues; see
+  `domains/messaging/delivery-semantics.md` for the correctness
+  properties of each semantic and the dead-letter strategy
+  comparison.
+
+## Observability — Consumer Lag (P10 Messaging is Observable)
+
+- Consumer lag (messages enqueued minus messages acked, or the
+  age of the oldest unacked message) is the primary queue health
+  signal. A lag that grows without bound is a P6 violation (the
+  consumer is slower than the producer) and a P10 violation if it
+  is not alerted.
+- DLQ depth is the secondary signal: a DLQ that grows is a P5
+  signal (poison messages are accumulating) and a P10 signal if
+  not alerted. Wire both to `domains/observability/metrics` as
+  SLIs with SLOs (e.g., lag < 1000 messages, DLQ depth < 10).
+- A queue with no lag metric is operating blind (P10 violation):
+  the operator cannot see the consumer falling behind until the
+  downstream effect surfaces.
+
+## Cross-Link to Concurrency (P6, cross-link concurrency/patterns)
+
+- The broker-backed queue is the cross-process analog of the
+  in-process bounded buffer. `domains/concurrency/patterns`
+  Pattern 5 (Bounded Queue with Backpressure) owns the in-process
+  instance ("producer is blocked or signaled" within one program);
+  messaging owns the broker-backed instance above it (the producer
+  is the broker's enqueue, the consumer is the broker's dequeue, the
+  backpressure is the prefetch cap and the lag signal). The
+  failure model differs: in-process fails by OOM; broker-backed
+  fails by network partition, broker restart, or consumer
+  crash-and-retry (D-062).
+- The cross-link is one-directional outward (messaging →
+  concurrency) per D-026 extended: messaging references concurrency
+  as the in-process foundation; concurrency does not back-link to
+  messaging.
+
+## What Violates Queue Discipline
+
+| Violation | Principle |
+|-----------|-----------|
+| Schemaless message (no versioned contract; consumer parses by guess) | P1 Messages are Contracts |
+| Non-idempotent consumer under at-least-once delivery (redelivery doubles the effect) | P3 Consumers are Idempotent |
+| Unstated delivery semantic (at-least-once vs exactly-once guessed) | P4 Delivery Semantics are Explicit |
+| No DLQ (poison message retried forever or silently dropped) | P5 Dead-Letter Handling is Defined |
+| Unbounded prefetch (broker floods consumer memory) | P6 Backpressure is Bounded |
+| Unbounded retry budget (no cap; consumer never progresses past the poison) | P5, P6 |
+| Consumer that assumes FIFO on a standard queue (ordering not guaranteed) | P2 Ordering is a Property, Not an Assumption |
+| Visibility timeout shorter than processing time (redeliver while still processing) | P4, P6 |
+| DLQ with no depth alert (poison messages accumulate invisibly) | P10, `domains/observability/metrics` |
+| No consumer-lag metric (consumer falls behind invisibly) | P10, `domains/observability/metrics` |
+| Dedup store that grows without bound (memory leak) | P6 Backpressure is Bounded |
+| Default prefetch with no rationale (underutilizes or OOMs) | P6, `domains/concurrency/patterns` |
@@ -0,0 +1,329 @@
+# Streams — Derived Rules
+
+> Derives from `domains/messaging/first-principles.md`. Applies P8
+> (Replay and Retention are Configured) primarily, with P2
+> (per-partition ordering), P3 (idempotent consumers), P4 (exactly-
+> once via transactions), P7 (partitioning), and P10 (observable
+> consumer-group health). For the Kafka / Kinesis / Pulsar /
+> NATS JetStream decision, see the stream-platform comparison
+> table below. For the consumer-group rebalance strategy choice,
+> see the rebalance enumeration below (IDEATE-41). Cross-links
+> `domains/messaging/delivery-semantics` for exactly-once via
+> transactions, `domains/data/schema-design` for stream schema,
+> and `domains/observability/tracing` for cross-partition traces.
+
+## What a Stream Is (P8 Replay and Retention are Configured)
+
+- A stream is a durable-log messaging primitive. Messages are
+  appended to a partitioned, replicated log; consumers read from an
+  offset and advance at their own pace. The log is retained for a
+  configured window (P8) — a stream is a durable log, not a pipe.
+  A consumer that falls behind can catch up by replaying from an
+  earlier offset; a consumer that starts fresh can replay from the
+  beginning (within retention).
+- The boundary with queues and pub/sub is the durability model. A
+  queue deletes on ack; a pub/sub subscription retains only its
+  unacked window; a stream retains the whole log for the configured
+  retention. This makes a stream replayable (P8 — the
+  reversibility mechanism) and observable as a record (P10 — the
+  log is itself an audit of what happened). See
+  `domains/messaging/pubsub.md` for the pub/sub-vs-stream
+  durability boundary discussion.
+- The boundary with concurrency is per D-062: messaging owns the
+  cross-process/network-failure-model angle. A stream is a
+  messaging concern because the log spans brokers and consumers
+  across a network, and the failure model is partition, broker
+  restart, or consumer crash-and-retry — not in-process OOM or
+  thread crash.
+
+## Partitioned Log Model (P2, P7)
+
+- A stream is partitioned for throughput and parallelism. Each
+  partition is an ordered, append-only log; messages within a
+  partition are strictly ordered (P2 — per-partition strict order
+  is the documented property). Across partitions, there is no
+  ordering guarantee: partition 0 and partition 1 are independent
+  logs.
+- The partition key (P7 — Partitioning is Intentional) determines
+  which partition a message lands on. A key that hashes evenly
+  spreads load; a key that matches the per-entity ordering need
+  (e.g., `userId` for user events) keeps a user's events on one
+  partition in order; a key that is too coarse (one partition for
+  the whole topic) serializes all traffic.
+- The partition count is a capacity bound: it caps the parallelism
+  (one consumer per partition per consumer group) and the
+  throughput (each partition has a write-throughput limit). Too
+  few partitions cap parallelism; too many partition overhead the
+  broker (file handles, replication, rebalance cost). The choice
+  is documented (P7), not defaulted.
+
+## Offsets (P2, P8)
+
+- An offset is a consumer's position in a partition. The consumer
+  reads from its last committed offset; acking (committing the
+  offset) advances it. A consumer that crashes before committing
+  re-reads from the last committed offset (at-least-once by
+  default, P4) — the consumer must be idempotent (P3).
+- The offset is per-partition (P2): each partition has its own
+  position, and the consumer commits them independently (or
+  atomically across partitions in a transaction — see below).
+- Replay (P8) is resetting the offset backward: a consumer can
+  replay from the beginning of retention, from a timestamp, or
+  from a specific offset. This is the durable-log property that
+  distinguishes a stream from a queue.
+
+## Consumer Groups (P3, P7, P10)
+
+- A consumer group is a set of consumers sharing the stream's
+  partitions: each partition is assigned to exactly one consumer
+  in the group. The group is the unit of parallelism and the unit
+  of offset tracking. Within a group, each consumer handles its
+  assigned partitions; across groups, each group independently
+  reads the whole stream (the pub/sub fan-out property, per
+  subscription/group).
+- A consumer in the group is idempotent (P3): under at-least-once
+  (the default), a redelivery after a crash-and-retry re-processes
+  messages. The consumer dedups by idempotency key, or processes
+  deterministically (e.g., a stateful aggregation that overwrites
+  with the latest value).
+- The consumer group's health is observable (P10): per-partition
+  lag (offset of the consumer vs the log's head), the group's
+  consumption rate, and rebalance events are first-class signals.
+  A group whose lag grows without bound is a P6 (backpressure)
+  smell and a P10 (observability) violation if not alerted.
+
+```python
+# Consumer-group reading from offsets (P2 per-partition order,
+# P3 idempotent under at-least-once, P7 partition assignment,
+# P8 replay from offset). Each consumer in the group handles its
+# assigned partitions; the group commits offsets atomically or
+# per-partition.
+
+def consume_stream(stream, group, dedup_store, process_event):
+    # Assign partitions to this consumer by the group's
+    # rebalance strategy (see the enumeration below).
+    for partition in stream.assigned_partitions(group, consumer=ME):
+        # Read from the last committed offset (P8 — replay by
+        # resetting this offset).
+        offset = stream.committed_offset(group, partition)
+        for message in stream.read(partition, from_offset=offset):
+            # P3: idempotent under at-least-once. A redelivery
+            # after a crash-and-retry re-processes; dedup by key.
+            if dedup_store.seen(message["idempotencyKey"]):
+                stream.commit(group, partition, message["offset"])
+                continue
+            process_event(message["payload"])
+            dedup_store.mark(message["idempotencyKey"])
+            # Commit the offset to advance (P8 — the position is
+            # the replay pointer).
+            stream.commit(group, partition, message["offset"])
+```
+
+- The commit-after-process order gives at-least-once (a crash
+  before commit re-reads); the commit-before-process order gives
+  at-most-once (a crash after commit loses the unprocessed
+  message). The default is at-least-once with idempotent consumers
+  (P3, P4).
+
+## Consumer-Group Rebalance Strategies (IDEATE-41, ATELIER-100 refinement)
+
+When a consumer joins or leaves the group, the broker must
+reassign partitions. The rebalance strategy determines the cost
+and the use-case fit. This enumeration parallels the v0.3
+IDEATE-30 drift-type enumeration (each strategy with its
+stop-the-world cost and use-case fit).
+
+| Strategy | Mechanism | Partition Stop-the-World Cost | Use-Case Fit |
+|----------|-----------|-------------------------------|--------------|
+| **Eager rebalance** (stop-the-world) | Every consumer in the group revokes ALL its partitions, the broker reassigns the full partition set, then consumers resume. Every rebalance pauses the whole group. | High — every partition pauses for every rebalance; the whole group stops processing during the revocation+reassignment window. Throughput drops to zero during rebalance. | Simple brokers, small groups, or rarely-rebalancing groups where the simplicity of full revocation outweighs the pause cost. Kafka's legacy protocol (pre-2.4). Avoid for large groups or frequent scale events. |
+| **Sticky (incremental cooperative) rebalance** | The broker reassigns only the partitions that must move (the joining/leaving consumer's share); existing partitions stay assigned. The rebalance is incremental and cooperative — no full revocation. | Low — only the moving partitions pause; the rest of the group continues processing. The pause is proportional to the changed partition count, not the total. | The default for large groups, frequent scale events, and rolling deploys. Kafka's CooperativeStickyAssignor (2.4+), Pulsar, NATS JetStream. Prefer for any group where a full stop-the-world on every deploy is unacceptable. |
+| **Cooperative (no-revoke) rebalance** | A subset of sticky where no partition is revoked unless the consumer leaves; only additions are incremental. The strictest minimization of stop-the-world. | Lowest — only added partitions pause; existing assignments are untouched. | Groups where partition assignment is append-only (consumers join but rarely leave). Useful for long-lived consumers with incremental scaling. |
+
+- The default for any non-trivial group is **sticky/cooperative**:
+  a rolling deploy that triggers an eager rebalance pauses the
+  whole group on every pod restart, which is unacceptable at
+  scale. The eager strategy is a legacy default that survives
+  because it is simple; prefer sticky where the broker supports
+  it.
+- The "partition stop-the-world cost" column is the P6
+  (backpressure) angle: a full stop-the-world during rebalance
+  causes lag to spike (the consumer is paused, the producer is
+  not). Sticky rebalance bounds the spike to the moving partitions.
+- A rebalance that pauses without a lag alert is a P10 violation:
+  the operator cannot see the rebalance-induced lag. Wire rebalance
+  events to `domains/observability/metrics` as an event signal.
+
+## Replay and Retention Windows (P8, C5 Reversibility)
+
+- Retention is the configured window the log keeps messages: time-
+  based (e.g., 7 days), size-based (e.g., 10 GB per partition), or
+  compacted (keep the latest value per key — a changelog). A
+  stream with no retention is a pipe, not a log (P8 violation); a
+  stream with infinite retention grows until disk exhaustion (P6
+  violation — backpressure on the broker).
+- Replay (P8) is resetting a consumer's offset to re-read from
+  within the retention window. Use cases: reprocessing after a
+  consumer bug (replay from the timestamp of the buggy deploy),
+  backfilling a new consumer (replay from the beginning), or
+  reindexing (replay to rebuild a derived store).
+- Compacted topics (Kafka log-compaction, Pulsar compaction) keep
+  the latest value per key and discard older values for the same
+  key. This turns the log into a changelog — a durable
+  materialized view that replays to the current state. Compaction
+  is a P8 (retention) and C5 (reversibility) mechanism: the log
+  retains the current state per key and is replayable to it.
+
+## Stream Processing (P2, P3, P4)
+
+- Stream processing is computing over the stream as it arrives:
+  windowing (tumbling, sliding, session windows), joins (stream-
+  stream, stream-table), aggregations (count, sum, per-key
+  windows), and stateful transformations. The processing is
+  per-partition ordered (P2 — a window over a partition is
+  deterministic; a window across partitions is not unless the
+  window is global).
+- Stream processing consumers are idempotent (P3): a redelivery
+  after a crash re-processes a window; the aggregation must
+  tolerate re-application (e.g., a sum is idempotent under replay
+  if the window is keyed by offset range, not by wall time).
+- Exactly-once stream processing (P4) requires transactions: the
+  consume-process-produce loop is one transaction — the input
+  offset commit and the output produce are atomic. See the
+  transactional exactly-once producer below.
+
+## Exactly-Once via Transactions (P4, P3)
+
+- Exactly-once stream processing is at-least-once plus a
+  transaction: the consumer commits the input offset and produces
+  the output in one transactional operation. If the consumer
+  crashes mid-transaction, neither the offset commit nor the
+  output produce happens — the consumer re-reads from the last
+  committed offset and re-processes (at-least-once), but the
+  transaction ensures the output is produced exactly once.
+- This is NOT a broker guarantee of exactly-once delivery; it is
+  at-least-once delivery plus idempotent/transactional processing
+  (P3, P4). Jepsen analyses of Kafka transactions confirm the
+  boundaries: the transaction is atomic within the broker, but
+  the downstream sink must be transactional or idempotent too.
+- See `domains/messaging/delivery-semantics.md` for the full
+  exactly-once-via-idempotency discussion.
+
+```python
+# Transactional exactly-once producer (P4 exactly-once via
+# transactions, P3 idempotent produce). The consume-process-
+# produce loop is one transaction: the input offset commit and
+# the output produce are atomic. A crash mid-transaction rolls
+# both back; the consumer re-reads and re-processes.
+
+def consume_transform_produce(stream, group, txn_producer):
+    # Begin a transaction. All produces and the offset commit in
+    # this block are atomic (P4).
+    with txn_producer.transaction() as txn:
+        for partition in stream.assigned_partitions(group, ME):
+            offset = stream.committed_offset(group, partition)
+            for message in stream.read(partition, from_offset=offset):
+                output = transform(message["payload"])
+                # P3: idempotent produce. The txn producer
+                # dedups by an epoch+sequence so a retried
+                # transaction does not double-produce.
+                txn.produce(
+                    topic="enriched-events",
+                    key=message["key"],
+                    value=output,
+                    idempotencyKey=message["idempotencyKey"],
+                )
+                # Commit the input offset within the same
+                # transaction (P4 atomicity). A crash before
+                # txn.commit() rolls this back; the consumer
+                # re-reads from the prior offset.
+                txn.commit_offset(group, partition, message["offset"])
+        # txn.commit() makes the produces and the offset commit
+        # visible atomically. A crash before this point aborts
+        # both; a crash after is safe (idempotent produce — P3).
+```
+
+- The transactional producer's idempotency (P3) is the defense
+  against a retried transaction: the broker dedups the output by
+  the producer's epoch and sequence so a re-commit does not
+  double-produce. The transaction (P4) is the defense against a
+  partial failure: the offset and the output commit together.
+
+## Stream Schema (P1, P9, cross-link data/schema-design)
+
+- A stream's messages carry a versioned schema (P1). The schema
+  evolves compatibly (P9): a new field the old consumer ignores is
+  backward-compatible; a renamed field the old consumer parses as
+  `undefined` is a P1 violation compounded as an evolution defect.
+- Stream schemas are often registered in a schema registry
+  (Confluent, Apicurio) that enforces compatibility on produce.
+  A producer that tries to publish an incompatible schema is
+  rejected; the registry is the P1/P9 enforcement point.
+- See `domains/data/schema-design` for the generic schema-design
+  discipline (Avro, Protobuf, JSON Schema); messaging owns the
+  stream-specific instance — the registry, the per-topic
+  compatibility mode, the consumer-side routing by schema
+  version.
+
+## Stream-Platform Comparison (D-069)
+
+| Axis | Apache Kafka | AWS Kinesis | Apache Pulsar | NATS JetStream |
+|------|--------------|-------------|---------------|----------------|
+| **Ordering** | Per-partition strict (P2); global only via single-partition topic | Per-shard strict; global only via single shard | Per-partition strict; global via single partition; also supports shared (out-of-order) subscriptions | Per-stream strict; per-subject ordering; global via single stream |
+| **Partitioning model** | Partitions (immutable count post-creation; increase requires recreate); key→partition by hash | Shards (reshardable: split/merge at runtime); key→shard by hash | Partitions (resizable; Pulsar's layered architecture separates compute from storage); key→partition by hash | Streams (subject-based; republish to resize); key→stream by subject |
+| **Replay / retention** | Time- or size-based retention; compaction (latest-per-key); replay from offset or timestamp | Time-based retention (24h–365d); replay from sequence number or timestamp; no compaction | Time- or size-based; compaction; replay from offset or timestamp; tiered storage (hot/warm/cold) | Time- or size-based; per-stream max-age; replay from sequence; no native compaction |
+| **Consumer groups** | Group-coordinated; offsets stored in an internal topic; eager (legacy) or sticky/cooperative (2.4+) rebalance | Enhanced fan-out consumers (per-shard HTTP/2 push); KCL for group coordination; no native group rebalance (shard is the unit) | Group-coordinated; shared or failover subscription modes; cooperative rebalance | Per-stream consumers; durable cursors; no native group rebalance (stream is the unit) |
+| **Exactly-once** | Transactions (KIP-98): consume-process-produce atomic; idempotent producer (KIP-516) | No native exactly-once; at-least-once with consumer-side dedup (P3) | Transactions: produce-ack atomic; idempotent producer | At-least-once by default; dedup window per stream (P3 idempotency) |
+| **Use-case fit** | High-throughput durable logs, stream processing (Kafka Streams, Flink), event sourcing, multi-consumer replay | AWS-native streaming, log ingestion, simple ETL within AWS; low operational burden | Cloud-native, geo-replication, tiered storage, mixed pub/sub + streaming; multi-tenant | Lightweight, low-latency, edge-friendly; NATS ecosystem; simpler ops than Kafka |
+| **Watch out for** | Partition count is fixed at creation (resize requires recreate + republish); rebalance cost on large groups; operational complexity | Shard limits per account; no compaction; retention cap at 365 days; AWS lock-in | Two-arch (BookKeeper + Brokers) operational complexity; smaller ecosystem | Smaller ecosystem; no native compaction; fewer stream-processing libraries |
+
+- The default for high-throughput durable logs with multi-consumer
+  replay is **Kafka**; for AWS-native streaming, **Kinesis**; for
+  geo-replicated multi-tenant or mixed pub/sub + streaming,
+  **Pulsar**; for lightweight low-latency edge-friendly streaming,
+  **NATS JetStream** (which also cross-links `domains/edge/iot`
+  via MQTT parallels — see the edge↔messaging bidirectionality in
+  `domains/messaging/first-principles.md` §4).
+- The ordering column is the P2 check: every platform provides
+  per-partition/per-shard strict order; none provides global order
+  across partitions except by single-partition. The replay/
+  retention column is the P8 check: every platform retains for a
+  configured window; replay is from offset or timestamp. The
+  exactly-once column is the P4 check: Kafka and Pulsar provide
+  transactions; Kinesis and JetStream rely on at-least-once plus
+  consumer-side idempotency (P3).
+
+## Cross-Partition Traces (P10, cross-link observability/tracing)
+
+- A stream-processing pipeline that fans out across partitions
+  must propagate a trace context per event: the trace ID follows
+  the event from source to processed output, even as the event
+  crosses partition boundaries. Without cross-partion traces, a
+  downstream error cannot be traced back to its source event.
+- See `domains/observability/tracing` for the generic distributed-
+  tracing discipline (trace context propagation, span
+  correlation). Messaging owns the stream-specific instance: the
+  trace context is a message header, the span boundary is the
+  consume-process-produce edge, and the cross-partition
+  correlation is by trace ID (not by partition — partitions are
+  independent logs, P2).
+- A stream processor with no trace propagation is a P10
+  violation: the operator cannot trace a processed event back to
+  its source. Wire the trace context into every produce and every
+  consume.
+
+## What Violates Stream Discipline
+
+| Violation | Principle |
+|-----------|-----------|
+| Stream with no retention (pipe, not log; no replay) | P8 Replay and Retention are Configured |
+| Infinite retention (grows until disk exhaustion) | P8, P6 |
+| Consumer group with eager rebalance at scale (full stop-the-world per deploy) | P6, IDEATE-41 (use sticky/cooperative) |
+| Non-idempotent stream consumer under at-least-once (redelivery re-processes the window) | P3 Consumers are Idempotent |
+| Default partition key (no rationale; hotspot or wrong-order) | P7 Partitioning is Intentional |
+| Partition count too low (caps parallelism) or too high (overhead) | P7 |
+| Cross-partition order assumption (no global order guarantee) | P2 Ordering is a Property, Not an Assumption |
+| Exactly-once claimed without transactional consume-process-produce (P4 violation) | P4 Delivery Semantics are Explicit |
+| Stream schema with no registry / no compatibility enforcement (silent shape break) | P1, P9, `domains/data/schema-design` |
+| No per-partition lag metric (consumer falls behind invisibly) | P10, `domains/observability/metrics` |
+| No cross-partition trace propagation (downstream error untraceable) | P10, `domains/observability/tracing` |
+| Compacted topic treated as a full log (old values already discarded) | P8, C1 (compaction is a retention mode, not a full log) |
@@ -0,0 +1,171 @@
+# Go Concurrency — Derived Application
+
+> Applies Atelier's domain principles to Go's concurrency specifically. Go's distinctive strength (goroutines, channels, context) earns a dedicated concurrency doc rather than a `go-async.md`.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/go.md` for the language first-principles stub.
+
+## Goroutines and Structured Concurrency (Concurrency P1 Immutability by Default, C6 Composability)
+
+- **`go f()` spawns a goroutine; ensure it does not outlive its parent:** an unstructured `go f()` leaks when the parent returns. Use `sync.WaitGroup`, `errgroup.Group`, or a `context`-scoped pattern to bound lifetime.
+- **`errgroup.WithContext` for structured concurrency:** a `Group` cancels its context on first error; siblings see the cancellation and exit. Mirrors `TaskGroup` semantics cross-language.
+- **Goroutines share only immutable inputs:** `go process(snap)` where `snap` is a copy. A goroutine sharing a mutable slice with the parent is a race (Concurrency P1 Immutability, P6 No Silent Races).
+- **No `go` in a library function without a documented lifetime:** a library that spawns unbounded goroutines leaks them into the caller. Either accept a `context.Context` or return a `Stop()` method.
+
+```go
+import "golang.org/x/sync/errgroup"
+
+func fetchAll(ctx context.Context, ids []string) ([]*User, error) {
+    g, ctx := errgroup.WithContext(ctx)
+    results := make([]*User, len(ids))
+    for i, id := range ids {
+        i, id := i, id  // capture loop vars
+        g.Go(func() error {
+            u, err := fetchUser(ctx, id)
+            if err != nil { return err }
+            results[i] = u
+            return nil
+        })
+    }
+    if err := g.Wait(); err != nil {
+        return nil, err
+    }
+    return results, nil
+}
+```
+
+## Channels: Bounded Queues and Backpressure (Concurrency P9 Bounded Queues, C6 Composability)
+
+- **Bounded channels apply backpressure:** `make(chan T, N)` blocks the sender when full (Concurrency P9 — bounded queues). Unbounded `make(chan T)` lets the producer run ahead and OOM.
+- **`select` with `default` for non-blocking send/receive:** a `default` case makes the channel a queue with try semantics; without it, the operation blocks.
+- **Close channel from the sender, never the receiver:** closing a channel signals "no more sends." A receiver closing it is a race; the sender may still be writing.
+- **One channel, one responsibility:** do not multiplex control and data on the same channel. Use a `select` over multiple channels instead.
+- **Applies `messaging/queues`:** a bounded Go channel is an in-process broker — bounded buffer, backpressure, at-most-once handoff. The same semantics apply; the broker is local.
+
+```go
+func pipeline(ctx context.Context, in <-chan Job, out chan<- Result) {
+    for {
+        select {
+        case j, ok := <-in:
+            if !ok { return }
+            r := process(j)
+            select {
+            case out <- r:
+            case <-ctx.Done():
+                return
+            }
+        case <-ctx.Done():
+            return
+        }
+    }
+}
+
+// bounded: backpressure when out is full
+out := make(chan Result, 16)
+```
+
+## context.Context for Cancellation (Concurrency P7 Cancellation Support, Concurrency P8 Timeout Discipline)
+
+- **`context.Context` is the first parameter of every I/O function:** `func fetchUser(ctx context.Context, id string) (*User, error)`. A function that does I/O without a `ctx` cannot be cancelled (Concurrency P7).
+- **`context.WithTimeout` for a deadline:** `ctx, cancel := context.WithTimeout(ctx, 5*time.Second); defer cancel()`. Every external call races against a deadline (Concurrency P8).
+- **`cancel()` always called, even on success:** `defer cancel()` immediately after creating the context. A leaked context leaks its timer.
+- **Never store a `context.Context` in a struct:** pass it as a parameter. A struct holding a `ctx` captures a request-scoped value into a long-lived object.
+- **Applies `concurrency/P7`:** cancellation propagates via `ctx.Done()`. A `select` on `<-ctx.Done()` is the cancel-aware wait.
+
+```go
+func fetchWithTimeout(ctx context.Context, url string) (*Response, error) {
+    ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
+    defer cancel()
+
+    req, _ := http.NewRequestWithContext(ctx, "GET", url, nil)
+    resp, err := http.DefaultClient.Do(req)
+    if err != nil {
+        if errors.Is(err, context.DeadlineExceeded) {
+            return nil, ErrTimeout
+        }
+        return nil, err
+    }
+    return resp, nil
+}
+```
+
+## select and Multiplexed Channels (Concurrency P7 Cancellation Support, C6 Composability)
+
+- **`select` multiplexes channel operations:** it picks a ready case at random (fair). A `select` with `<-ctx.Done()` plus a data case is the cancel-aware wait.
+- **`default` makes `select` non-blocking:** use for "send if ready, else drop" (a bounded queue with drop-oldest policy).
+- **`select {}` blocks forever:** a `select{}` with no cases is a permanent block. Use only in a goroutine that should run until the process exits.
+- **Applies `concurrency/P7`:** the `select` over `ctx.Done()` and a result channel is the canonical cancel pattern.
+
+```go
+func processUntilCancel(ctx context.Context, jobs <-chan Job) {
+    for {
+        select {
+        case <-ctx.Done():
+            return
+        case j, ok := <-jobs:
+            if !ok { return }
+            // ...
+        }
+    }
+}
+```
+
+## sync Primitives and Lock Scope (Concurrency P3 Boundaries are Locks, Concurrency P5 Lock Minimization)
+
+- **`sync.Mutex` scoped minimally:** not held across I/O (a `Send` on a channel, an HTTP call). Hold the lock, mutate, release — then do I/O (Concurrency P3 Lock Scope).
+- **`sync.RWMutex` for read-heavy, `Mutex` for write-heavy:** RWMutex adds overhead; only prefer it when reads dominate by 10x+.
+- **`sync.Map` for specific cases (append-only, disjoint keys):** not a general `map[K]V` replacement. For most maps, `Mutex` + `map` is clearer and often faster.
+- **`sync.Once` for one-time init:** `var once sync.Once; once.Do(func(){ init() })`. Idempotent and race-free.
+- **Applies `concurrency/P5` (lock minimization):** prefer channels over locks; when a lock is needed, hold it for the smallest possible scope.
+
+```go
+type Cache struct {
+    mu    sync.Mutex
+    items map[string]*User
+}
+
+func (c *Cache) Get(id string) (*User, bool) {
+    c.mu.Lock()
+    defer c.mu.Unlock()
+    u, ok := c.items[id]
+    return u, ok
+}
+
+func (c *Cache) Set(id string, u *User) {
+    c.mu.Lock()
+    c.items[id] = u
+    c.mu.Unlock()   // explicit unlock before any I/O
+}
+```
+
+## Race Detection (Concurrency P6 No Silent Races)
+
+- **`go test -race` enforces `P6`:** the race detector instruments memory accesses and fails on data races. See `go-tooling.md` for the CI gate.
+- **Tests must exercise the concurrent path:** a serial test of a `Mutex`-protected map finds no race. Write tests with N goroutines hitting the map under `-race`.
+- **Applies `concurrency/P6`:** a race detected at test time is a bug fixed; a race undetected is a production heisenbug.
+
+```go
+func TestCacheConcurrent(t *testing.T) {
+    c := &Cache{items: map[string]*User{}}
+    var wg sync.WaitGroup
+    for i := 0; i < 100; i++ {
+        i := i
+        wg.Add(1)
+        go func() {
+            defer wg.Done()
+            c.Set(strconv.Itoa(i), &User{})
+            c.Get(strconv.Itoa(i))
+        }()
+    }
+    wg.Wait()
+}
+```
+
+## Cross-References
+
+- `domains/concurrency/patterns.md` — the cancellation/timeout/semaphore patterns applied here.
+- `domains/concurrency/first-principles.md` — Concurrency P1, P3, P5, P6, P7, P8, P9 traced throughout.
+- `domains/messaging/queues.md` — bounded Go channels as in-process brokers; backpressure parallels (IDEATE-40).
+- `domains/errors/patterns.md` — `errgroup` and error propagation in concurrent code.
+- `languages/go-types.md` — typed channels carry the named types defined there.
+- `languages/go-tooling.md` — the `-race` CI gate that enforces Concurrency P6.
+- `languages/go-testing.md` — concurrent tests that exercise the race detector.
@@ -0,0 +1,141 @@
+# Go Testing — Derived Application
+
+> Applies Atelier's domain principles to Go testing specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/go.md` for the language first-principles stub.
+
+## Table-Driven Tests (Testing P1 Tests as Specification, C2 Clarity)
+
+- **Table-driven is the Go idiom:** `cases := []struct{ name string; in X; want Y }{...}`; loop with `t.Run(c.name, ...)`. Each case is a subtest with its own name and failure output.
+- **Test names read as a spec:** `{"rejects empty email", ...}`, `{"returns persisted id", ...}`. A reader understands the unit from the subtest names (Testing P1).
+- **No `if got != want { t.Fatal() }` shared across cases:** each case asserts independently; a failure in case 3 does not skip cases 4 and 5.
+- **`t.Run` enables `-run` filtering:** `go test -run TestCreateUser/rejects_empty_email` runs one case. Essential for debugging a single failure.
+
+```go
+func TestCreateUser(t *testing.T) {
+    cases := []struct {
+        name    string
+        email   string
+        wantErr bool
+    }{
+        {"rejects empty email", "", true},
+        {"rejects missing @", "no-at-sign", true},
+        {"accepts valid email", "a@b.co", false},
+    }
+    for _, c := range cases {
+        t.Run(c.name, func(t *testing.T) {
+            _, err := CreateUser(c.email)
+            if (err != nil) != c.wantErr {
+                t.Fatalf("err=%v, wantErr=%v", err, c.wantErr)
+            }
+        })
+    }
+}
+```
+
+## t.Parallel for Independence (Testing P2 Independence, Concurrency P10 Test for Race Conditions)
+
+- **`t.Parallel()` for independent subtests:** each subtest opts in; the runner executes them concurrently. A test that fails under `Parallel` has hidden state (Testing P2 Independence).
+- **Capture loop variables:** `c := c` inside the loop, or rely on Go 1.22+ per-iteration scoping. A parallel subtest sharing `c` races on the last value.
+- **Applies `concurrency/P10` (test for races):** parallel tests are the first line of race detection; combine with `-race` for the full safety net.
+
+```go
+for _, c := range cases {
+    c := c  // capture for parallel
+    t.Run(c.name, func(t *testing.T) {
+        t.Parallel()
+        _, err := CreateUser(c.email)
+        if (err != nil) != c.wantErr {
+            t.Fatalf("err=%v, wantErr=%v", err, c.wantErr)
+        }
+    })
+}
+```
+
+## t.Cleanup for Teardown (Testing P3 Determinism, Testing P2 Independence)
+
+- **`t.Cleanup(func() { ... })` for teardown:** runs in LIFO order after the test (and its subtests) complete. Replaces `defer` in a helper that does not know when the test ends.
+- **Per-test state, not shared:** a `setup(t)` helper creates resources and registers cleanup; each test gets its own. A package-level `var` shared across tests is order coupling.
+- **`t.TempDir()` for filesystem tests:** creates a unique temp dir and cleans up automatically. No manual `os.RemoveAll` and no cross-test contamination.
+- **Applies `Testing P3` (determinism):** cleanup is tied to the test lifecycle, not a global teardown that may run before or after depending on order.
+
+```go
+func setupStore(t *testing.T) *Store {
+    t.Parallel()
+    dir := t.TempDir()  // auto-cleaned
+    s, err := OpenStore(filepath.Join(dir, "db"))
+    if err != nil { t.Fatal(err) }
+    t.Cleanup(func() { s.Close() })
+    return s
+}
+```
+
+## Race Detector (Testing P9 Edge Case Coverage, Concurrency P6 No Silent Races)
+
+- **`go test -race` in CI, always:** see `go-tooling.md`. The detector is the enforcement of `concurrency/P6`.
+- **Tests must exercise the concurrent path:** a serial test of a `Mutex`-protected map finds no race. Write tests with N goroutines.
+- **`-count=1` to disable result caching:** by default, Go caches passing tests. `-count=1` forces re-run; combine with `-race` and parallelism to surface heisenbugs.
+- **Applies `Testing P9` (edge case coverage):** the race detector is the edge-case tool for concurrency — it finds the inputs the test author forgot to write.
+
+```bash
+# CI gate
+go test -race -count=1 ./...
+```
+
+## Time and Determinism (Testing P3 Determinism, Testing P9 Edge Case Coverage)
+
+- **No `time.Now()` in code under test:** inject a `Clock` interface. In tests, a fake clock advances deterministically.
+- **`time.Sleep` in tests is a smell:** a sleep waits for a real timer, flaky under load. Use a channel or `Eventually`-style polling with a timeout.
+- **`t.Deadline()` aware helpers:** a helper that may take long checks `t.Deadline()` and bails early. Prevents a slow test from timing out the suite.
+
+```go
+type Clock interface { Now() time.Time }
+
+type fakeClock struct{ t time.Time }
+func (f *fakeClock) Now() time.Time { return f.t }
+
+func TestUserHasCreatedAt(t *testing.T) {
+    clk := &fakeClock{time.Date(2024, 1, 1, 0, 0, 0, 0, time.UTC)}
+    u, _ := CreateUserWithClock("a@b.co", clk)
+    if u.CreatedAt.Year() != 2024 {
+        t.Fatalf("year=%d, want 2024", u.CreatedAt.Year())
+    }
+}
+```
+
+## Mocks and Interfaces (Testing P7 Realism, API P1 Contract Fidelity)
+
+- **Mock at the interface, not the struct:** `type Store interface { Get(id string) (*User, error) }` in production; `type mockStore struct{ ... }` in test. The interface is the contract (applies `api/P1`).
+- **`httptest` for HTTP servers:** `httptest.NewServer` gives a real server on a loopback port; no manual socket plumbing.
+- **`testify/mock` or hand-written mocks:** hand-written for one-off, `testify` for complex sequencing. Avoid mocking frameworks that generate code at runtime (reflection-heavy) — they hide failures behind stack traces.
+- **Applies `Testing P7` (realism):** mock the boundary (HTTP, DB), not the unit. Mocking the unit under test tests the mock.
+
+```go
+type mockStore struct {
+    users map[string]*User
+    got  []string
+}
+func (m *mockStore) Get(id string) (*User, error) {
+    m.got = append(m.got, id)
+    return m.users[id], nil
+}
+
+func TestGetUserLogs(t *testing.T) {
+    s := &mockStore{users: map[string]*User{"abc": {}}}
+    svc := NewService(s)
+    svc.GetUser("abc")
+    if len(s.got) != 1 || s.got[0] != "abc" {
+        t.Fatalf("got=%v", s.got)
+    }
+}
+```
+
+## Cross-References
+
+- `domains/testing/pyramid.md` — where unit/integration/race tests sit; the race job is its own layer.
+- `domains/testing/fixtures.md` — `t.TempDir` and `t.Cleanup` as the fixture discipline.
+- `domains/testing/first-principles.md` — Testing P1 Specification, P2 Independence, P3 Determinism, P9 Edge Coverage.
+- `domains/concurrency/first-principles.md` — Concurrency P6 (race detector), P10 (test for races).
+- `languages/go-types.md` — the named types tests assert.
+- `languages/go-concurrency.md` — concurrent tests exercise the patterns from that doc.
+- `languages/go-tooling.md` — the `go test` flags (`-race`, `-count`, `-run`) detailed here.
@@ -0,0 +1,97 @@
+# Go Tooling — Derived Application
+
+> Applies Atelier's domain principles to Go tooling specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/go.md` for the language first-principles stub.
+
+## go vet and golangci-lint (DevOps P2 Automation, C2 Clarity)
+
+- **`go vet` is the stdlib baseline:** it catches `printf` format mismatches, lock-copy-by-value, and unreachable code. Run on every build.
+- **`golangci-lint` aggregates vet + dozens of linters:** enable `errcheck` (no `_ = err`), `govet`, `staticcheck`, `ineffassign`, `unused`, `gofmt`, `goimports`. Each enabled linter has a one-line `# reason:` in `.golangci.yml`.
+- **`errcheck` enforces `errors/P2` (fail loudly):** a discarded error is a silent failure. `errcheck` fails the build on `_ = doX()`.
+- **`goimports` over `gofmt`:** `goimports` adds missing imports and removes unused ones, in addition to formatting. The format is not debated in review (Clarity C2).
+
+```yaml
+# .golangci.yml
+linters:
+  enable:
+    - errcheck    # reason: Errors P2 — no swallowed errors
+    - govet
+    - staticcheck
+    - ineffassign
+    - unused
+    - gofmt
+    - goimports
+linters-settings:
+  errcheck:
+    check-blank: true   # fail on _ = fn()
+```
+
+## go test -race (Concurrency P6 No Silent Races)
+
+- **`go test -race` in CI, always:** the race detector instruments memory accesses and fails on data races. It is the primary enforcement of `concurrency/P6` (no silent races).
+- **`-race` adds overhead; run it in a separate CI job:** the race build is ~2x slower; keep the fast unit-test job and add a race job.
+- **`-race` requires tests that actually exercise the concurrent path:** a test that calls `Get`/`Set` serially finds no race. Write tests that spawn goroutines hitting the same map.
+- **Applies `concurrency/P6`:** a race detected is a bug fixed; a race undetected is a heisenbug in production. The detector is the safety net.
+
+```bash
+# CI race job
+go test -race -count=1 ./...
+```
+
+## Module Discipline (DevOps P1 Reproducibility)
+
+- **`go mod tidy` on every change that touches imports:** removes unused deps and adds missing ones. A `go.mod` with stale entries breaks reproducibility.
+- **`go.sum` committed and verified:** `go mod verify` checks the checksums of the module cache against `go.sum`. A drifted `go.sum` is a supply-chain signal.
+- **Pinned major versions in `go.mod`:** `require github.com/x/y v1.2.3` pins the minor; a `v1.2.4` patch may auto-update. For applications, consider a `go.mod` proxy that pins to exact commits.
+- **`go mod vendor` for hermetic CI:** vendoring `vendor/` into the repo means CI builds without network. The trade-off is repo size; the win is reproducibility (DevOps P1).
+
+```bash
+# CI build gate
+go mod tidy
+go mod verify
+go build ./...
+go test -race ./...
+```
+
+## Reproducible Builds (DevOps P1 Reproducibility, C3 Simplicity)
+
+- **One Go toolchain version, pinned:** `goenv` or `asdf` pins the Go version per repo; a `.go-version` file declares it. A CI job that uses "latest" Go drifts.
+- **`CGO_ENABLED=0` for static binaries:** a static binary runs in a scratch container with no libc dependency. Set in CI for all release builds.
+- **`-trimpath` and `-ldflags='-s -w'` for reproducible output:** strips the build path from the binary and removes debug info. Two builds of the same commit produce byte-identical binaries.
+
+```bash
+# Reproducible release build
+CGO_ENABLED=0 go build -trimpath -ldflags='-s -w' -o app ./cmd/app
+```
+
+## Documentation in the Pipeline (Documentation P1 Documentation is Code, DevOps P9 Documentation in the Pipeline)
+
+- **`go doc` from comments:** package comments and exported-symbol comments are the API docs; `go doc` and `pkg.go.dev` render them. Missing comments on exported symbols fail `revive`/`golint` (Documentation P1).
+- **`// Example` functions are run by `go test`:** an `ExampleUser` function with `// Output:` is a tested artifact; a stale output fails the build.
+- **`README.md` and `docs/` are built by `mkdocs` or similar:** the pipeline validates links and renders; a broken link fails CI (Documentation P1).
+
+```go
+// GetUser fetches a user by id.
+//
+// Example:
+//
+//	u, err := GetUser(id)
+//	if err != nil { ... }
+func GetUser(id UserId) (*User, error) { /* ... */ }
+
+func ExampleGetUser() {
+    u, err := GetUser("abc")
+    fmt.Println(u, err)
+    // Output: <nil> not found
+}
+```
+
+## Cross-References
+
+- `domains/devops/ci-cd.md` — the pipeline gates that host vet/lint/test.
+- `domains/devops/first-principles.md` — DevOps P1 Reproducibility, P2 Automation.
+- `domains/concurrency/first-principles.md` — Concurrency P6 No Silent Races (`-race`).
+- `domains/documentation/first-principles.md` — Documentation P1 Documentation is Code.
+- `languages/go-types.md` — the type rules staticcheck enforces reference this doc.
+- `languages/go-testing.md` — the `go test` flags (`-race`, `-count`) detailed here.
@@ -0,0 +1,140 @@
+# Go Type System — Derived Application
+
+> Applies Atelier's domain principles to Go's type system specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/go.md` for the language first-principles stub.
+
+## Named Types for Domain Concepts (C1 Correctness, Data P7 Type Fidelity)
+
+- **Named types for domain IDs and values:** `type UserId string`, `type OrderId string`. Two named types are distinct even with identical underlying types; the compiler rejects the swap.
+- **Constructors validate at the boundary:** `func NewUserId(s string) (UserId, error)` returns an error on bad input. A bare `UserId(s)` cast bypasses validation — only the constructor is exported.
+- **Applies `data/P7` (type fidelity):** a named type carries the domain meaning through the call graph; a `string` parameter does not.
+- **`any` is the wide type; narrow before use:** Go 1.18+ `any` is an alias for `interface{}`. Use it only at true boundaries (e.g., `json.Unmarshal`); narrow with a type assertion immediately.
+
+```go
+type UserId string
+type OrderId string
+
+func NewUserId(s string) (UserId, error) {
+    if !regexp.MustCompile(`^[a-z0-9]+$`).MatchString(s) {
+        return "", fmt.Errorf("invalid user id: %q", s)
+    }
+    return UserId(s), nil
+}
+
+func GetUser(id UserId) (*User, error) { /* ... */ }
+
+// GetUser("abc")      // compile error: string is not UserId
+// GetUser(OrderId("abc")) // compile error: distinct named types
+```
+
+## Generics (C6 Composability, Data P7 Type Fidelity)
+
+- **Generics (1.18+) preserve element types across containers:** `type Repository[T any] struct { ... }` keeps `T` through `Get`/`Save`, rather than widening to `any`.
+- **Constrain with `comparable` for map keys, custom interfaces for behavior:** `func dedupe[T comparable](s []T) []T` uses `comparable`; a `Sortable[T]` constraint expresses the `Less` requirement.
+- **Avoid generics where an interface suffices:** `io.Reader` is not improved by generics. Generics are for type-preserving containers; interfaces are for behavior.
+- **No generic methods on generic types (not supported):** `func (r Repository[T]) Map[U any](f func(T) U) Repository[U]` is a compile error. Use a free function.
+
+```go
+type Entity interface { ID() string }
+
+type Repository[T Entity] struct {
+    db map[string]T
+}
+
+func (r *Repository[T]) Get(id string) (T, bool) {
+    var zero T
+    t, ok := r.db[id]
+    if !ok { return zero, false }
+    return t, true
+}
+
+func (r *Repository[T]) Save(t T) { r.db[t.ID()] = t }
+```
+
+## Interfaces (C6 Composability, API P1 Contract Fidelity)
+
+- **Interfaces defined by the consumer, not the producer:** a package defines its dependencies as interfaces (`type Store interface { Get(id string) (*User, error) }`), and accepts implementations. The producer does not pre-declare "the interface I implement."
+- **Small interfaces (Go proverb):** `io.Reader` is one method. An interface with 5+ methods is a god-object; split it.
+- **Accept interfaces, return concrete types:** return a `*UserRepo`, accept a `Store`. The caller gets the implementation; the callee depends on the abstraction.
+- **Applies `api/P1` (contract fidelity):** the interface is the contract; the concrete type is the implementation. Tests mock the interface, not the struct.
+
+```go
+// consumer defines the interface
+type UserStore interface {
+    Get(id string) (*User, error)
+}
+
+type Service struct { store UserStore }
+
+func NewService(s UserStore) *Service { return &Service{store: s} }
+
+// producer returns concrete; satisfies UserStore implicitly
+type UserRepo struct { db map[string]*User }
+func (r *UserRepo) Get(id string) (*User, error) { return r.db[id], nil }
+```
+
+## Type Assertion Discipline (C1 Correctness, Errors P1 Errors are Data)
+
+- **Type assertions return `(T, bool)` — use the bool:** `v, ok := x.(UserId)` distinguishes "wrong type" from "zero value." A bare `x.(UserId)` panics on mismatch.
+- **`switch x := x.(type)` for multi-variant narrowing:** each case narrows `x` to the case type. The default case is exhaustive (no `never`-style check; Go relies on review).
+- **Applies `errors/P1` (errors are data):** a failed type assertion is a value (`ok == false`), not an exception. Handle it as a branch, not a panic.
+- **Never assert across module boundaries silently:** an assertion on a type from another package couples to its internals. Prefer an interface method.
+
+```go
+func describe(x any) string {
+    switch v := x.(type) {
+    case UserId:
+        return "user " + string(v)
+    case OrderId:
+        return "order " + string(v)
+    default:
+        return fmt.Sprintf("unknown: %T", v)
+    }
+}
+
+// safe form, never panic
+id, ok := raw.(UserId)
+if !ok {
+    return fmt.Errorf("expected UserId, got %T", raw)
+}
+```
+
+## Error Types and errors.Is/As (Errors P1 Errors are Data, Errors P3 Fail Specifically)
+
+- **Sentinel errors for known cases:** `var ErrNotFound = errors.New("not found")`; check with `errors.Is(err, ErrNotFound)`. The sentinel is a value, not an exception class.
+- **Custom error types for context:** `type ValidationError struct { Field, Msg string }`; check with `var ve *ValidationError; errors.As(err, &ve)`. The type carries structured data (Errors P4 Preserve Context).
+- **Wrap with `%w`:** `fmt.Errorf("get user %s: %w", id, err)` preserves the chain. `errors.Is`/`As` unwrap it. Bare `%v` breaks the chain.
+- **Applies `errors/P3` (fail specifically):** `ErrNotFound` is specific; `ErrFailed` is not. The error type names the failure mode.
+
+```go
+var ErrNotFound = errors.New("not found")
+
+type ValidationError struct {
+    Field string
+    Msg   string
+}
+func (e *ValidationError) Error() string { return e.Field + ": " + e.Msg }
+
+func GetUser(id UserId) (*User, error) {
+    u, ok := db[string(id)]
+    if !ok {
+        return nil, fmt.Errorf("user %s: %w", id, ErrNotFound)
+    }
+    return u, nil
+}
+
+// caller
+if errors.Is(err, ErrNotFound) { /* 404 */ }
+var ve *ValidationError
+if errors.As(err, &ve) { /* 422 with ve.Field */ }
+```
+
+## Cross-References
+
+- `domains/data/schema-design.md` — named types parallel schema design at the Go boundary.
+- `domains/data/first-principles.md` — Data P7 Type Fidelity is the primary trace.
+- `domains/api/rest.md` — contract fidelity for HTTP handlers using interfaces.
+- `domains/errors/patterns.md` — `errors.Is`/`As` and the wrap-with-`%w` pattern.
+- `languages/go-concurrency.md` — typed channels carry these named types.
+- `languages/go-testing.md` — table-driven tests assert type-swap safety.
@@ -2,6 +2,13 @@

 > How Atelier's domain principles apply in Go specifically. Derives from `domains/` docs.

+## Derived Docs
+
+- [go-types.md](go-types.md) — named types, generics, interfaces, type assertion discipline.
+- [go-tooling.md](go-tooling.md) — go vet, golangci-lint, go test -race, module discipline.
+- [go-concurrency.md](go-concurrency.md) — goroutines, channels, context, select, sync primitives.
+- [go-testing.md](go-testing.md) — table-driven tests, t.Parallel, t.Cleanup, race detector.
+
 ## Type System (C1 Correctness, Data P7 Type Fidelity)

 - **Named types for domain concepts:** `type UserId string`, not bare `string`.
@@ -0,0 +1,117 @@
+# Python Async — Derived Application
+
+> Applies Atelier's domain principles to Python async specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/python.md` for the language first-principles stub.
+
+## asyncio and anyio (Concurrency P7 Cancellation Support, C2 Clarity)
+
+- **`asyncio` for I/O-bound work; threads only for blocking libraries:** `async def` + `await` for network/disk; `run_in_executor` to wrap a blocking call. Mixing threads for I/O is the wrong default.
+- **`anyio` for runtime portability:** `anyio` abstracts asyncio/trio; a library written against `anyio` runs on either. Use it for libraries; for applications, asyncio directly is fine.
+- **One event loop, one thread:** `asyncio.run(main())` creates and runs the loop. Do not call `asyncio.run` inside an existing loop (raises `RuntimeError`); do not share a loop across threads.
+- **Applies `concurrency/P7`:** every `async def` accepts cancellation as a first-class signal; `CancelledError` propagates unless explicitly suppressed (and suppressing it is almost always a bug).
+
+```python
+import asyncio
+import anyio
+
+async def fetch_user(id: str) -> User:
+    return await api.get(f'/users/{id}')
+
+# asyncio application
+async def main():
+    user = await fetch_user('abc')
+
+asyncio.run(main())
+
+# anyio library — portable across asyncio/trio
+async def fetch_all(ids: list[str]) -> list[User]:
+    return await anyio.gather(*(fetch_user(i) for i in ids))
+```
+
+## Structured Concurrency (Concurrency P1 Immutability by Default, C6 Composability)
+
+- **`asyncio.TaskGroup` (3.11+) for structured concurrency:** tasks created in a `TaskGroup` are awaited or cancelled together on exit. No orphan tasks outlive the block.
+- **No `asyncio.gather(..., return_exceptions=False)` for fallible tasks:** `gather` returns partial results on first exception; `TaskGroup` cancels siblings and propagates the error atomically. Use `TaskGroup` for new code.
+- **Applies `concurrency/P1` (immutability):** tasks share only immutable inputs; results are collected, not mutated in place. A task that writes to a shared list is a race waiting to happen.
+- **`anyio.create_task_group()` mirrors `TaskGroup` cross-runtime:** same structured-concurrency guarantee, portable.
+
+```python
+import asyncio
+
+async def fetch_all(ids: list[str]) -> list[User]:
+    results: list[User] = []
+    async with asyncio.TaskGroup() as tg:
+        tasks = [tg.create_task(fetch_user(i)) for i in ids]
+    # all tasks done (or cancelled) by here
+    return [t.result() for t in tasks]
+```
+
+## Cancellation and Timeout (Concurrency P7 Cancellation Support, Concurrency P8 Timeout Discipline)
+
+- **`asyncio.wait_for(coro, timeout)` for a deadline:** every external `await` races against a timeout. A bare `await` is an unbounded wait (Concurrency P8).
+- **`asyncio.timeout()` (3.11+) as a context manager:** `async with asyncio.timeout(5): await op` — cleaner than `wait_for` for multi-await blocks.
+- **`CancelledError` propagates; do not catch broadly:** `except Exception` swallows `CancelledError` in 3.7 (it was `BaseException`); in 3.8+ it's `BaseException` and `except Exception` skips it. Catch specifically, never bare `except:`.
+- **Applies `concurrency/P7`:** cancellation is cooperative — a long synchronous block inside `async def` ignores cancellation. Yield with `await asyncio.sleep(0)` periodically in CPU-bound loops.
+
+```python
+import asyncio
+
+async def fetch_with_timeout(id: str, timeout: float = 5.0) -> User:
+    async with asyncio.timeout(timeout):
+        return await fetch_user(id)
+
+async def shutdown(token: asyncio.Event) -> None:
+    # cooperative cancel — long-running loop checks the token
+    while not token.is_set():
+        await do_chunk()
+        await asyncio.sleep(0)   # yield so cancel can land
+```
+
+## Bounded Concurrency and Queues (Concurrency P9 Bounded Queues)
+
+- **`asyncio.Semaphore(N)` to bound in-flight tasks:** a `Semaphore(8)` wrapping `gather` caps concurrency. Unbounded `gather` on a 10k-item list exhausts file descriptors (Concurrency P9 — bounded queues).
+- **`asyncio.Queue(maxsize=N)` for producer/consumer:** a bounded queue applies backpressure to the producer. An unbounded queue lets the producer run ahead and OOM.
+- **Applies `messaging/queues`:** an `asyncio.Queue` is an in-process broker — the same bounded-queue / backpressure semantics apply; the broker is just local.
+
+```python
+import asyncio
+
+async def map_bounded(items: list[str], limit: int = 8) -> list[User]:
+    sem = asyncio.Semaphore(limit)
+    async def guarded(i: str) -> User:
+        async with sem:
+            return await fetch_user(i)
+    return await asyncio.gather(*(guarded(i) for i in items))
+```
+
+## Error Handling in Async (Errors P5 Recoverable When Possible, Errors P1 Errors are Data)
+
+- **Retry with backoff for transient failures:** network blips are recoverable (Errors P5). Exponential backoff with jitter, capped retries, and an `anyio`-cancellation-aware `sleep`.
+- **No retry for non-idempotent operations:** a `POST` creating a resource is not safely retryable without an idempotency key (applies `api/P6` Idempotency).
+- **`except asyncio.CancelledError: raise`** is the only valid handling — re-raise so the cancellation propagates. Catching and continuing breaks structured concurrency.
+- **Applies `messaging/delivery-semantics`:** a cancelable async operation is at-most-once; retry-on-cancel is at-least-once. The caller must declare which.
+
+```python
+import anyio
+import random
+
+async def fetch_retry(id: str, attempts: int = 3) -> User:
+    for i in range(attempts):
+        try:
+            return await fetch_user(id)
+        except (TimeoutError, ConnectionError):
+            if i == attempts - 1:
+                raise
+            await anyio.sleep((2 ** i) * 0.1 + random.random() * 0.1)
+    raise RuntimeError('unreachable')
+```
+
+## Cross-References
+
+- `domains/concurrency/patterns.md` — the cancellation/timeout/semaphore patterns applied here.
+- `domains/concurrency/first-principles.md` — Concurrency P1 Immutability, P7 Cancellation Support, P8 Timeout Discipline, P9 Bounded Queues.
+- `domains/messaging/queues.md` — `asyncio.Queue` as an in-process broker; backpressure parallels (IDEATE-40).
+- `domains/errors/patterns.md` — typed async errors and retry-with-backoff.
+- `languages/py-types.md` — `Result` and exception hierarchy used in async error handling.
+- `languages/py-tooling.md` — `pytest-asyncio` config that runs these tests.
@@ -0,0 +1,111 @@
+# Python Testing — Derived Application
+
+> Applies Atelier's domain principles to Python testing specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/python.md` for the language first-principles stub.
+
+## pytest and Spec-Driven Tests (Testing P1 Tests as Specification, C2 Clarity)
+
+- **`pytest` is the default; `unittest` only for stdlib-only libraries:** `pytest` fixtures, parametrize, and assertion rewriting beat `unittest`'s boilerplate (Clarity C2).
+- **Tests co-located with source:** `user.py` → `test_user.py`. A test far from its subject rots (Documentation P5 Discoverability).
+- **Test names read as a spec:** `def test_create_user_rejects_invalid_email():` — a reader understands the unit from the name. Avoid `def test_user1():`.
+- **`assert` over `self.assertEqual`:** pytest rewrites `assert` to show the failing values; `assertEqual` is unittest's escape hatch and loses readability.
+- **Applies `Testing P1`:** the test is a specification; the failure message is the spec violation.
+
+```python
+# test_user.py
+import pytest
+from user import create_user, ValidationError
+
+def test_create_user_rejects_invalid_email():
+    with pytest.raises(ValidationError):
+        create_user(email='not-an-email')
+
+def test_create_user_returns_persisted_id():
+    u = create_user(email='a@b.co')
+    assert u.id  # truthy persisted id
+```
+
+## Factories and Fixture Discipline (Testing P2 Independence, Testing P7 Realism)
+
+- **`factory_boy` or `pytest-factoryboy` over shared fixtures for mutable state:** `UserFactory.build()` returns a fresh object per call; a session-scoped fixture mutated across tests couples them (Testing P2 Independence).
+- **Fixtures for setup/teardown, factories for data:** a `db` fixture sets up the DB once per test; a `make_user` factory produces fresh data per assertion. Conflating them produces order-dependent tests.
+- **`scope='function'` is the default and the safe default:** `scope='session'` for read-only resources (a schema migration), never for mutable state.
+- **Mock at the boundary, not the unit:** `mocker.patch('requests.get')` for HTTP; do not patch `user.User.save` (that mocks the unit under test — Testing P7 realism).
+
+```python
+import factory
+from user import User
+
+class UserFactory(factory.Factory):
+    class Meta:
+        model = User
+    email = factory.Sequence(lambda n: f'u{n}@b.co')
+    name = 'Test User'
+
+def test_user_factory_is_fresh():
+    u1 = UserFactory.build()
+    u2 = UserFactory.build()
+    assert u1.email != u2.email   # independent
+```
+
+## Parametrize and Edge Cases (Testing P9 Edge Case Coverage, Testing P3 Determinism)
+
+- **`@pytest.mark.parametrize` for input tables:** one parametrized test runs N cases; each is an independent test with its own name and failure output (Testing P9).
+- **Edge cases as rows, not special tests:** empty list, `None`, max int, unicode — each a row. An ad-hoc `test_handles_edge` with multiple asserts hides which case failed (Testing P6 Failure Specificity).
+- **`pytest --randomly` catches order coupling:** a test passing alone but failing in a suite has hidden shared state. The random plugin makes it visible (Testing P2 Independence).
+- **Property tests via `hypothesis`:** for invariants (e.g., "parse(serialize(x)) == x"), `hypothesis` generates hundreds of inputs and shrinks failures to a minimal counterexample.
+
+```python
+import pytest
+
+@pytest.mark.parametrize('email, reason', [
+    ('', 'empty'),
+    ('a' * 1000 + '@b.co', 'too long'),
+    ('no-at-sign', 'missing @'),
+    ('a@b', 'missing TLD'),
+])
+def test_create_user_rejects(email, reason):
+    with pytest.raises(ValidationError):
+        create_user(email=email)
+```
+
+## Determinism and Time (Testing P3 Determinism, Testing P9 Edge Case Coverage)
+
+- **No `datetime.now()`, `time.time()`, `uuid.uuid4()`, `random.random()` in code under test:** inject a `Clock`, `UUIDGen`, `Random` port. In tests, provide deterministic fakes.
+- **`freezegun` for time:** `@freeze_time('2024-01-01')` makes `datetime.now()` deterministic. Do not call `datetime.now()` directly in code — wrap it in a `Clock` port so production and tests both inject.
+- **`pytest --randomly-seed=last` to reproduce a failing order:** when `--randomly` finds an order bug, the seed is logged; re-run with it to debug deterministically.
+
+```python
+from freezegun import freeze_time
+
+@freeze_time('2024-01-01')
+def test_user_has_created_at():
+    u = create_user(email='a@b.co')
+    assert u.created_at.year == 2024
+```
+
+## Async Tests (Concurrency P10 Test for Race Conditions, Testing P1 Tests as Specification)
+
+- **`pytest-asyncio` (or `anyio`'s pytest plugin) for `async def` tests:** `@pytest.mark.asyncio` runs the coroutine on a loop. Without it, an `async def` test is silently skipped (returns a coroutine, never awaited).
+- **`anyio`'s plugin runs the same test on asyncio and trio:** one parametrized run across both runtimes catches runtime-specific bugs.
+- **Race-sensitive tests use `--randomly` and bounded concurrency:** a `Semaphore(1)` test under random order surfaces hidden state.
+- **Applies `concurrency/P10`:** async tests are the race detector's first line — if a test passes alone but fails under `gather` of N, there's a race.
+
+```python
+import pytest
+
+@pytest.mark.asyncio
+async def test_async_fetch_returns_user():
+    u = await fetch_user('abc')
+    assert u.email
+```
+
+## Cross-References
+
+- `domains/testing/pyramid.md` — where unit/integration/property tests sit; hypothesis is the property layer.
+- `domains/testing/fixtures.md` — factory-vs-fixture discipline applied via `factory_boy`.
+- `domains/testing/first-principles.md` — Testing P1 Specification, P2 Independence, P3 Determinism, P9 Edge Coverage.
+- `languages/py-types.md` — the `Result` and Pydantic models that tests assert.
+- `languages/py-async.md` — async tests use the cancellation/timeout patterns from that doc.
+- `languages/py-tooling.md` — the `pyproject.toml [tool.pytest]` config that runs these tests.
@@ -0,0 +1,99 @@
+# Python Tooling — Derived Application
+
+> Applies Atelier's domain principles to Python tooling specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/python.md` for the language first-principles stub.
+
+## ruff for Lint and Format (DevOps P2 Automation, C2 Clarity)
+
+- **`ruff` replaces flake8 + black + isort + pyupgrade:** one tool, one config, one order of magnitude faster. Format is not debated in review (Clarity C2).
+- **Rule selection is principled, not "everything":** `select = ["E", "F", "I", "UP", "B", "SIM"]` — each rule group has a one-line `# reason:` in `pyproject.toml`. Rules without a rationale are noise (Documentation P1 — docs are code).
+- **`ruff format` is the formatter, `ruff check` is the linter:** run both in CI; the formatter is deterministic, the linter surfaces smells.
+- **Applies `devops/P2`:** the format/lint gate runs on every push; a developer never waits for a reviewer to comment on style.
+
+```toml
+# pyproject.toml
+[tool.ruff]
+target-version = "py311"
+line-length = 100
+
+[tool.ruff.lint]
+select = ["E", "F", "I", "UP", "B", "SIM", "RUF"]
+# reason: E/F = pyflakes+pycodestyle; I = isort; UP = pyupgrade; B = bugbear; SIM = simplification
+
+[tool.ruff.format]
+quote-style = "double"
+```
+
+## mypy and Type-Check Gate (DevOps P2 Automation, Data P7 Type Fidelity)
+
+- **`mypy --strict` in CI, not in the editor:** strict flags (`disallow_untyped_defs`, `no_implicit_optional`, `warn_return_any`) are the floor. The editor runs a relaxed mypy for speed; CI runs strict as the gate.
+- **`pyright` for stricter/async-aware checking:** pyright understands `async` better and reports faster; mypy is the standard. Pick one as the gate, run the other as informational.
+- **Per-module overrides only with a tracked reason:** `[[tool.mypy.overrides]] module = "legacy.*" ignore_errors = true` — each override block links to a ticket. Untracked overrides accumulate into a permanently untyped core.
+- **`py.typed` marker for libraries:** ships the type info to consumers. Without it, downstream mypy treats the library as `Any`.
+
+```bash
+# CI gate
+mypy --strict src/
+pyright src/ || true   # informational
+```
+
+## Dependency Management: poetry and uv (DevOps P1 Reproducibility)
+
+- **`poetry` or `uv` for lockfile discipline:** both produce a deterministic lock (`poetry.lock` / `uv.lock`). `pip install` alone does not — it resolves at install time, producing different trees across machines.
+- **`uv` for speed (Rust-based, 10–100x faster):** newer tool, same lockfile semantics. Either is acceptable; do not mix within a repo.
+- **Lockfile committed for applications:** for libraries, commit the lock for CI reproducibility even though consumers resolve their own tree.
+- **`--frozen` install in CI:** `poetry install --no-dev --frozen` fails if the lock is out of sync. Prevents a "works on my machine" drift.
+
+```bash
+# CI install — deterministic
+uv sync --frozen --no-dev
+# or
+poetry install --no-dev --frozen
+```
+
+## Virtualenv Discipline (DevOps P1 Reproducibility, C3 Simplicity)
+
+- **One virtualenv per project, never the system Python:** `uv venv` or `python -m venv .venv`. System Python drift breaks reproducibility.
+- **`uv` creates and pins the Python version:** `uv venv --python 3.12` ensures the same interpreter across machines. A pinned Python is part of the reproducibility contract, not just the lockfile.
+- **No `pip install` into the system Python in CI:** use `uv`/`poetry`'s venv. A CI step that mutates system Python makes the next job non-hermetic.
+
+```bash
+uv venv --python 3.12
+source .venv/bin/activate
+uv pip install -r requirements.txt
+```
+
+## Documentation in the Pipeline (Documentation P1 Documentation is Code, DevOps P9 Documentation in the Pipeline)
+
+- **`mkdocs` + `mkdocstrings` from docstrings:** API docs are generated from `google`- or `numpy`-style docstrings; the build fails on missing docstrings for public symbols (Documentation P1).
+- **`doctest` blocks in docstrings are run by pytest:** a `>>>` example is a tested artifact; a stale example fails the build (Documentation P1, Testing P1).
+- **`pyproject.toml` is the single source of tool config:** ruff, mypy, pytest, poetry all read from it. Do not scatter `.flake8`, `setup.cfg`, `mypy.ini`. One config file is one place to look (Clarity C2).
+
+```python
+def get_user(id: UUID) -> User:
+    """Fetch a user by id.
+
+    Args:
+        id: the user's UUID.
+
+    Returns:
+        The User.
+
+    Raises:
+        NotFoundError: if the user does not exist.
+
+    Example:
+        >>> get_user(UUID('intentional-example-uuid'))
+        User(...)
+    """
+    ...
+```
+
+## Cross-References
+
+- `domains/devops/ci-cd.md` — the pipeline gates that host ruff/mypy/poetry.
+- `domains/devops/first-principles.md` — DevOps P1 Reproducibility, P2 Automation.
+- `domains/documentation/first-principles.md` — Documentation P1 Documentation is Code.
+- `languages/py-types.md` — the type rules mypy enforces reference this doc.
+- `languages/py-testing.md` — the pytest config (`pyproject.toml [tool.pytest]`) detailed here.
@@ -0,0 +1,111 @@
+# Python Type System — Derived Application
+
+> Applies Atelier's domain principles to Python's type system specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/python.md` for the language first-principles stub.
+
+## Type Hints and Gradual Typing (C1 Correctness, Data P7 Type Fidelity)
+
+- **Type hints on every function signature:** `def get_user(id: UUID) -> User | None:`. Hints are annotations, not enforcement, but `mypy`/`pyright` make them a build gate.
+- **Gradual typing is opt-in, not opt-out:** start with `--strict` on a package, fix the errors, then expand. A repo-wide `# type: ignore` is a gradual-typing failure.
+- **`from __future__ import annotations` for forward refs:** all annotations are strings until resolved, so `class User: ...` referencing `User` works without quotes in 3.10+.
+- **`Any` disables the checker; `object` is the wide type:** `Any` allows any operation; `object` requires narrowing. Use `object` for opaque inputs (e.g., `json.loads` return).
+- **Applies `data/P7` (type fidelity):** a hint is the contract; the checker verifies it. A missing hint is a missing contract.
+
+- **Pydantic for runtime validation at the boundary:** hints on `BaseModel` fields are validated at construction, catching bad input from network/config before it deepens into the system.
+- **`extra='forbid'` by default:** Pydantic allows extra fields silently; forbid them to surface schema drift (e.g., a client sending a typo'd field name).
+- **Custom types via `Annotated` with validators:** `Email = Annotated[str, validate_email]` keeps the type readable and the validator attached to the type, not the model.
+
+```python
+from pydantic import BaseModel, ConfigDict
+from uuid import UUID
+
+class UserCreate(BaseModel):
+    model_config = ConfigDict(extra='forbid')
+    email: str
+    name: str
+
+class User(UserCreate):
+    id: UUID
+```
+
+## Pydantic and Schema Fidelity (C1 Correctness, API P1 Contract Fidelity, Data P7 Type Fidelity)
+
+- **Pydantic models are the API contract:** a FastAPI handler taking `UserCreate` rejects malformed JSON with a 422 before the body runs. This is `api/P1` (contract fidelity) at the type boundary.
+- **`ConfigDict(extra='forbid')` rejects unknown fields:** silently accepting extras is a contract leak — the server appears to handle fields it ignores.
+- **Validators raise `ValueError`, not `Exception`:** Pydantic converts `ValueError` to a validation error response; a generic `Exception` becomes a 500 and hides the input bug.
+- **Applies `api/P1`:** the model is the source of truth; the OpenAPI schema is generated from it, not hand-written. Drift between schema and code is impossible.
+
+```python
+from typing import Annotated
+from pydantic import BaseModel, Field, StringConstraints
+
+EmailStr = Annotated[str, StringConstraints(pattern=r'^[^@\s]+@[^@\s]+$')]
+
+class Login(BaseModel):
+    email: EmailStr
+    password: Annotated[str, Field(min_length=8)]
+```
+
+## Errors as Data (Errors P1 Errors are Data, C1 Correctness)
+
+- **`Union[T, Error]` over `Optional[T]` for expected failures:** `Optional[User]` cannot distinguish "not found" from "permission denied". A discriminated `Result` carries the cause.
+- **Custom exception hierarchy rooted at `AppError`:** `class NotFoundError(AppError)` etc. — callers can `except AppError` for the broad case, or a specific subclass for handling.
+- **`raise` for exceptional paths, `return Result` for expected:** "user not found" is expected (a `Result`); "DB connection lost" is exceptional (a `raise`). Conflating them makes error handling a guess.
+- **Applies `errors/P1`:** errors are values, not control-flow magic. A `Result` type encodes this at the type level even where exceptions are the runtime mechanism.
+
+```python
+from dataclasses import dataclass
+from typing import Generic, TypeVar, Union
+
+T = TypeVar('T')
+
+@dataclass(frozen=True)
+class Ok(Generic[T]):
+    value: T
+
+@dataclass(frozen=True)
+class Err:
+    error: Exception
+
+Result = Union[Ok[T], Err]
+
+def find_user(id: UUID) -> Result[User]:
+    row = db.get(id)
+    if row is None:
+        return Err(NotFoundError(f'user {id}'))
+    return Ok(User.from_row(row))
+```
+
+## Generics and Protocols (C6 Composability, C1 Correctness)
+
+- **`Protocol` for structural typing (PEP 544):** a `Repository` protocol defines `get`/`save` without requiring an inheritance hierarchy; any class matching the shape satisfies it.
+- **`TypeVar` with bounds for generic functions:** `T = TypeVar('T', bound=Entity)` lets `serialize(t: T) -> dict` access `t.id`.
+- **`Generic[T]` for container types:** a typed `Repository[T]` preserves the element type across `get`/`save`, rather than widening to `Any`.
+- **`@overload` for callable overloads:** `def parse(s: str) -> int: ...` vs `def parse(s: bytes) -> int: ...` — the runtime body is one function; the overloads are the type contract.
+
+```python
+from typing import Protocol, TypeVar
+
+T = TypeVar('T')
+
+class Repository(Protocol[T]):
+    def get(self, id: str) -> T | None: ...
+    def save(self, t: T) -> None: ...
+
+class UserRepo:
+    def get(self, id: str) -> User | None: ...
+    def save(self, u: User) -> None: ...
+
+def use_repo(r: Repository[User]) -> None:
+    u = r.get('abc')   # type: User | None
+```
+
+## Cross-References
+
+- `domains/data/schema-design.md` — Pydantic models parallel schema design at the TS/JSON boundary.
+- `domains/data/first-principles.md` — Data P7 Type Fidelity is the primary trace for this doc.
+- `domains/api/rest.md` — contract fidelity for FastAPI handlers consuming Pydantic models.
+- `domains/errors/patterns.md` — the `Result` discriminated union as error-as-data encoding.
+- `languages/py-async.md` — typed async results built on the `Result` union here.
+- `languages/py-tooling.md` — the `mypy`/`pyright` config that enforces these hints.
@@ -2,6 +2,13 @@

 > How Atelier's domain principles apply in Python specifically. Derives from `domains/` docs.

+## Derived Docs
+
+- [py-types.md](py-types.md) — type hints + Pydantic, mypy/pyright, gradual typing.
+- [py-tooling.md](py-tooling.md) — ruff, mypy, poetry, uv, virtualenv discipline.
+- [py-async.md](py-async.md) — asyncio, anyio, cancellation, structured concurrency.
+- [py-testing.md](py-testing.md) — pytest, factory_boy, fixture discipline, parametrize.
+
 ## Type System (C1 Correctness, Data P7 Type Fidelity)

 - **Type hints on every function:** `def get_user(id: UUID) -> User | None:`.
@@ -0,0 +1,129 @@
+# Rust Async — Derived Application
+
+> Applies Atelier's domain principles to Rust async specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/rust.md` for the language first-principles stub.
+
+## tokio and the Async Runtime (Concurrency P5 Lock Minimization, C6 Composability)
+
+- **`tokio` is the default async runtime:** `#[tokio::main]` for the entry; `tokio::spawn` for a task. The runtime owns the reactor, the I/O driver, and the timer.
+- **`tokio::spawn` returns a `JoinHandle` like `std::thread::spawn`:** a dropped `JoinHandle` detaches (the task keeps running); `await` the handle to join. Prefer await to detach.
+- **`tokio::task::JoinSet` for structured concurrency:** a set of tasks awaited together; on drop, all remaining tasks are cancelled. Mirrors `errgroup`/`TaskGroup` semantics.
+- **`runtime` features are explicit:** `tokio = { version = "1", features = ["full"] }` for a binary; `["rt", "rt-multi-thread", "macros"]` for a library. Pulling `full` into a library bloats downstream.
+
+```rust
+#[tokio::main]
+async fn main() {
+    let mut set = tokio::task::JoinSet::new();
+    for id in ["a", "b", "c"] {
+        set.spawn(fetch_user(id.to_string()));
+    }
+    while let Some(res) = set.join_next().await {
+        match res {
+            Ok(Ok(u)) => println!("{}", u.name),
+            Ok(Err(e)) => eprintln!("err: {e}"),
+            Err(join_err) => eprintln!("panic: {join_err}"),
+        }
+    }
+}
+```
+
+## Async Traits (Concurrency P7 Cancellation Support, C6 Composability)
+
+- **`async fn` in traits stabilized in Rust 1.75:** `trait Repo { async fn get(&self, id: &str) -> Result<User, Error>; }`. No `async-trait` crate needed for new code on recent toolchains.
+- **`Box<dyn Trait>` with async methods needs `dyn`-compatibility:** the returned future is `Pin<Box<dyn Future>>`; the compiler boxes it. For hot paths, use generics (`impl Trait`) over `dyn`.
+- **`async-trait` crate for older toolchains:** macro that desugars to a `Pin<Box<dyn Future>>`. Migrate to native `async fn in trait` when the toolchain allows.
+- **`Send` bounds on async traits for cross-thread spawn:** `trait Repo: Send { async fn get(&self, id: &str) -> Result<User, Error>; }` — the returned future must be `Send` to spawn on a multi-thread runtime.
+
+```rust
+trait UserRepo: Send + Sync {
+    async fn get(&self, id: &str) -> Result<User, Error>;
+}
+
+struct PgRepo { pool: PgPool }
+impl UserRepo for PgRepo {
+    async fn get(&self, id: &str) -> Result<User, Error> {
+        sqlx::query_as::<_, User>("SELECT * FROM users WHERE id = $1")
+            .bind(id).fetch_one(&self.pool).await.map_err(Error::from)
+    }
+}
+```
+
+## Cancellation (Concurrency P7 Cancellation Support, Concurrency P8 Timeout Discipline)
+
+- **Cancellation is cooperative via dropping the future:** `tokio::select!` drops the unselected branch, cancelling it. A dropped future stops at its next `.await` point.
+- **`tokio::time::timeout` for a deadline:** `timeout(Duration::from_secs(5), op).await` returns `Ok(Ok(v))` on success, `Ok(Err(e))` on inner error, `Err(Elapsed)` on timeout. Every external `await` races against a deadline (Concurrency P8).
+- **`tokio::select!` for cancel-aware waits:** `select! { res = op => res, _ = cancel => return Err(Cancelled), }`. The unselected branch is dropped, cancelling it.
+- **Cancellation is not atomic:** a future dropped mid-`await` may have partial state. `Drop` runs on cancellation; clean up there (e.g., rollback a transaction).
+- **Applies `concurrency/P7`:** cancellation is a first-class signal; the runtime propagates it via drop. No `CancelledError` to catch — the future is gone.
+
+```rust
+use tokio::time::timeout;
+use std::time::Duration;
+
+async fn fetch_with_timeout(url: &str) -> Result<Response, Error> {
+    match timeout(Duration::from_secs(5), fetch(url)).await {
+        Ok(Ok(r)) => Ok(r),
+        Ok(Err(e)) => Err(e.into()),
+        Err(_elapsed) => Err(Error::Timeout),
+    }
+}
+
+async fn cancellable(op: impl Future<Output=()>, mut cancel: tokio::sync::oneshot::Receiver<()>) {
+    tokio::select! {
+        _ = op => {},
+        _ = &mut cancel => println!("cancelled"),
+    }
+}
+```
+
+## Pin and Self-Referential Futures (Concurrency P5 Lock Minimization, C1 Correctness)
+
+- **`async fn` returns a `Future` that is often self-referential:** the generated state machine may hold a borrow into its own stack. Such a future must be `Pin`ned to move safely.
+- **`Pin<Box<T>>` to box and pin:** `Box::pin(async { ... })` returns a `Pin<Box<dyn Future>>`. The cost is a heap alloc; the win is `Send`/`dyn`-compatibility.
+- **`Pin<&mut T>` for in-place polling:** `Pin::new(&mut fut)` pins a stack future; the borrow checker prevents moving it. Use for stack-allocated futures in `select!`.
+- **Do not `unsafe` unpin:** `Pin::get_unchecked_mut` opts out of the pin guarantees. Application code never needs it; library code uses it for `poll` implementations.
+
+```rust
+use std::pin::Pin;
+
+async fn boxed() -> Pin<Box<dyn std::future::Future<Output = ()> + Send>> {
+    Box::pin(async {
+        // self-referential state machine is safe to move once pinned
+    })
+}
+```
+
+## Bounded Channels and Backpressure (Concurrency P9 Bounded Queues)
+
+- **`tokio::sync::mpsc::channel(N)` is bounded:** `send().await` blocks when full (backpressure, Concurrency P9). Unbounded `unbounded_channel()` lets the producer run ahead and OOM.
+- **`tokio::sync::mpsc::Sender::try_send` for non-blocking send:** returns `Err(TrySendError::Full(v))` when full; the caller decides to drop, log, or back off. A bounded queue + `try_send` is the backpressure-aware pattern.
+- **`tokio::sync::broadcast` for fan-out:** multiple receivers each get a copy; a slow receiver misses (lag). Use for telemetry, not for commands.
+- **Applies `messaging/queues`:** a bounded tokio channel is an in-process broker — bounded buffer, backpressure, at-most-once handoff. The same semantics apply; the broker is local.
+
+```rust
+use tokio::sync::mpsc;
+
+async fn producer(tx: mpsc::Sender<Job>) {
+    for j in jobs() {
+        if tx.send(j).await.is_err() { return; } // receiver dropped
+    }
+}
+
+async fn consumer(rx: mpsc::Receiver<Job>) {
+    while let Some(j) = rx.recv().await {
+        process(j).await;
+    }
+}
+
+let (tx, rx) = mpsc::channel::<Job>(16);  // bounded: backpressure
+```
+
+## Cross-References
+
+- `domains/concurrency/patterns.md` — the cancellation/timeout/semaphore patterns applied here.
+- `domains/concurrency/first-principles.md` — Concurrency P5 Lock Minimization, P7 Cancellation Support, P8 Timeout Discipline, P9 Bounded Queues.
+- `domains/messaging/delivery-semantics.md` — at-most-once vs at-least-once framing for async retry/cancel (IDEATE-40).
+- `languages/rs-ownership.md` — `Send`/`Sync` bounds on futures build on the ownership model here.
+- `languages/rs-tooling.md` — `tokio` feature flags and the `cargo` build profiles detailed there.
+- `languages/rs-testing.md` — `#[tokio::test]` and async test patterns.
@@ -0,0 +1,136 @@
+# Rust Ownership — Derived Application
+
+> Applies Atelier's domain principles to Rust's ownership model specifically. Rust's distinctive strength (Send/Sync, lifetimes, borrowing) earns a dedicated ownership doc rather than an `rs-types.md`.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/rust.md` for the language first-principles stub.
+
+## Ownership and Move Semantics (Concurrency P1 Immutability by Default, C1 Correctness)
+
+- **Ownership is unique:** at any time, exactly one owner holds a value. Assignment passes ownership (`let y = x;` — `x` is moved, not copied). The compiler rejects use-after-move.
+- **`Copy` types (integers, `bool`, `&T`) duplicate on assignment; everything else moves.** A `struct` is `Copy` only if all fields are; opt in via `#[derive(Copy, Clone)]` only for small, cheap-to-copy types.
+- **Pass by `&T` for read-only, `&mut T` for mutation:** a borrow does not transfer ownership; the caller retains the value after the callee returns.
+- **Applies `concurrency/P1` (immutability by default):** `&T` is shared and immutable; `&mut T` is exclusive and mutable. The compiler enforces "one or many, never both" — aliasing XOR mutation, statically.
+
+```rust
+let s = String::from("hello");
+let t = s;          // s moved into t
+// println!("{}", s); // error: use of moved value
+
+let n = 5;
+let m = n;          // i32 is Copy: n still usable
+println!("{} {}", n, m);
+```
+
+## Borrowing and Lifetimes (C1 Correctness, Data P7 Type Fidelity, Concurrency P3 Boundaries are Locks)
+
+- **`&'a T` ties a borrow to a lifetime `'a`:** the borrow cannot outlive the owner. Lifetimes are static — the compiler rejects dangling references.
+- **Lifetime elision when unambiguous:** `fn first<'a>(s: &'a str) -> &'a str` is elided to `fn first(s: &str) -> &str` (one input → output lifetime). When ambiguous, name the lifetime.
+- **`'static` is the longest lifetime (the whole program):** not "until I drop it." Use `'static` only for values that genuinely live forever (string literals, `const`s); leaking to `'static` to satisfy the checker is a bug.
+- **`Ref<'a, T>` and `RefMut<'a, T>` from `RefCell` are runtime-checked borrows:** the borrow rules still apply, checked at runtime instead of compile time. A second `RefMut` panics.
+- **Applies `concurrency/P3` (boundaries are locks):** `&mut T` is the compile-time lock — exclusive access is the boundary; no runtime mutex needed for single-threaded aliasing discipline.
+
+```rust
+fn longest<'a>(a: &'a str, b: &'a str) -> &'a str {
+    if a.len() > b.len() { a } else { b }   // borrow tied to both inputs
+}
+
+fn dangling() -> &str {       // compile error: missing lifetime
+    let s = String::from("local");
+    &s                          // error: s drops at end of fn
+}
+```
+
+## Send and Sync (Concurrency P1 Immutability by Default, Concurrency P3 Boundaries are Locks, C1 Correctness)
+
+- **`Send`:** a type `T: Send` may be moved across thread boundaries. Most types are `Send`; `Rc<T>` is not (shared non-atomically refcounted).
+- **`Sync`:** a type `T: Sync` may be shared (`&T`) across threads. `RefCell<T>` is `!Sync` (interior mutability without atomics); `Mutex<T>` is `Sync` (it synchronizes).
+- **The compiler enforces `Send`/`Sync` at the thread-spawn boundary:** `std::thread::spawn(move || { ... })` requires the closure's captures to be `Send`.
+- **Applies `concurrency/P1` and `concurrency/P3`:** `Send` is the move-across-boundary contract; `Sync` is the share-across-boundary contract. Data races are a compile error, not a runtime detector. This is Rust's distinctive strength over Go's race detector.
+
+```rust
+use std::rc::Rc;
+use std::sync::Arc;
+
+let rc = Rc::new(5);
+// std::thread::spawn(move || { println!("{}", rc) }); // error: Rc is !Send
+
+let arc = Arc::new(5);
+std::thread::spawn(move || { println!("{}", arc) });    // ok: Arc<T> is Send+Sync
+```
+
+## Shared Mutation: Arc, Mutex, RwLock (Concurrency P3 Boundaries are Locks, Concurrency P5 Lock Minimization)
+
+- **`Arc<T>` for shared ownership across threads:** atomic refcounted. Clone increases the count; the last drop frees `T`.
+- **`Mutex<T>` for exclusive mutation across threads:** `lock()` blocks until exclusive; the guard `MutexGuard<T>` derefs to `&mut T` and releases on drop.
+- **`RwLock<T>` for read-heavy, `Mutex<T>` for write-heavy:** RwLock allows multiple readers or one writer. For most cases, `Mutex` is simpler and faster; prefer it unless reads dominate by 10x+.
+- **Hold the lock for the smallest scope:** `let g = m.lock().unwrap();` then drop `g` before I/O. RAII releases on scope exit; explicit `drop(g)` clarifies intent.
+- **Applies `concurrency/P5` (lock minimization):** prefer message passing (`mpsc` channels) over locks. When a lock is needed, scope it minimally.
+
+```rust
+use std::sync::{Arc, Mutex};
+use std::thread;
+
+let counter = Arc::new(Mutex::new(0));
+let mut handles = vec![];
+for _ in 0..10 {
+    let c = Arc::clone(&counter);
+    handles.push(thread::spawn(move || {
+        let mut g = c.lock().unwrap();
+        *g += 1;
+        // g drops here, lock released
+    }));
+}
+for h in handles { h.join().unwrap(); }
+println!("{}", *counter.lock().unwrap());
+```
+
+## Interior Mutability (Concurrency P1 Immutability by Default, C1 Correctness)
+
+- **`Cell<T>` for `Copy` types, `RefCell<T>` for non-`Copy`:** interior mutability moves the borrow check from compile time to runtime. `RefCell::borrow_mut()` panics on a second mutable borrow.
+- **`Mutex<T>`/`RwLock<T>` for thread-safe interior mutability:** the runtime check is the lock, not a panic. Use these across threads; `RefCell` only single-threaded.
+- **`UnsafeCell<T>` is the primitive; never use directly:** `Cell`, `RefCell`, `Mutex` are safe wrappers. Direct `UnsafeCell` is `unsafe` and opts out of the aliasing guarantee.
+- **Applies `concurrency/P1`:** interior mutability is the exception, not the default. Reach for it when an API must present `&self` while mutating internally (e.g., a cache); document why.
+
+```rust
+use std::cell::RefCell;
+
+struct Cache {
+    inner: RefCell<HashMap<String, User>>,
+}
+impl Cache {
+    fn get(&self, id: &str) -> Option<User> {
+        // &self (immutable) but mutates internally
+        self.inner.borrow_mut().entry(id.to_string()).or_insert_with(|| fetch()).clone()
+    }
+}
+```
+
+## Drop and RAII (C1 Correctness, Concurrency P3 Boundaries are Locks)
+
+- **`Drop` runs when the owner goes out of scope:** no `defer`, no `finally`. A `MutexGuard` releases, a `File` closes, a `JoinHandle`... does not join (a dropped `JoinHandle` detaches).
+- **`Drop` is deterministic:** it runs at scope exit, not GC time. This is why `Arc`'s refcount is precise and `Mutex` release is timely.
+- **`ManuallyDrop<T>` to opt out:** for FFI types whose destructor you must call manually. Rare in application code; common in `unsafe` bindings.
+- **`Drop` order: fields in declaration order, then the struct itself.** A field that another field's `Drop` depends on must be declared last.
+
+```rust
+struct Resource { name: String }
+impl Drop for Resource {
+    fn drop(&mut self) {
+        println!("dropping {}", self.name);   // runs at scope end
+    }
+}
+
+fn main() {
+    let _r = Resource { name: "x".into() };
+    // _r drops here, prints "dropping x"
+}
+```
+
+## Cross-References
+
+- `domains/concurrency/first-principles.md` — Concurrency P1 Immutability, P3 Boundaries are Locks, P5 Lock Minimization.
+- `domains/data/first-principles.md` — Data P7 Type Fidelity (lifetimes are the type-level fidelity for references).
+- `domains/concurrency/patterns.md` — message-passing vs lock patterns applied via `Arc`/`Mutex`/`mpsc`.
+- `domains/errors/patterns.md` — `?` propagation relies on ownership transfer of the error.
+- `languages/rs-async.md` — async borrows (`Pin`/`&mut`) build on the lifetime model here.
+- `languages/rs-testing.md` — `Send`/`Sync` tests and ownership-based property tests.
@@ -0,0 +1,159 @@
+# Rust Testing — Derived Application
+
+> Applies Atelier's domain principles to Rust testing specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/rust.md` for the language first-principles stub.
+
+## #[test] and Co-located Tests (Testing P1 Tests as Specification, C2 Clarity)
+
+- **`#[test]` on functions in a `#[cfg(test)] mod tests` block:** tests co-located with source, compiled only in `cargo test`. A test file far from its subject rots (Documentation P5 Discoverability).
+- **Test names read as a spec:** `fn create_user_rejects_invalid_email()` — a reader understands the unit from the name. Avoid `fn test_user_1()`.
+- **`assert!` / `assert_eq!` / `assert_ne!` over raw `panic!`:** the macros produce readable failure output (`assertion failed: left == right, left: 5, right: 3`). Raw `panic!` gives a message only (Testing P6 Failure Specificity).
+- **Applies `Testing P1`:** the test is a specification; the failure message is the spec violation.
+
+```rust
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn create_user_rejects_invalid_email() {
+        let r = create_user("not-an-email");
+        assert!(matches!(r, Err(Error::Validation(_))));
+    }
+
+    #[test]
+    fn create_user_returns_persisted_id() {
+        let u = create_user("a@b.co").unwrap();
+        assert!(!u.id.is_empty());
+    }
+}
+```
+
+## proptest and Property Tests (Testing P9 Edge Case Coverage, Testing P1 Tests as Specification)
+
+- **`proptest` (or `quickcheck`) for invariant tests:** declare a property (`parse(serialize(x)) == x`), the framework generates hundreds of inputs and shrinks failures to a minimal counterexample (Testing P9).
+- **Strategy over hand-written generators:** `proptest::collection::vec(any::<u32>(), 0..100)` generates arbitrary `Vec<u32>`; do not hand-roll a generator for each property.
+- **`proptest!` macro or `proptest! { ... }` block:** each `case (name) => { ... }` is a property. The block is the spec (Testing P1).
+- **Property tests complement, not replace, example tests:** examples document the happy path; properties cover the edge space. Both are required.
+
+```rust
+use proptest::prelude::*;
+
+proptest! {
+    #[test]
+    fn roundtrips_id(s in "[a-z0-9]{1,32}") {
+        let id = UserId::new(&s).unwrap();
+        assert_eq!(id.as_str(), s);
+    }
+
+    #[test]
+    fn rejects_invalid_id(s in "[^a-z0-9]+") {
+        assert!(UserId::new(&s).is_err());
+    }
+}
+```
+
+## Mock Discipline (Testing P2 Independence, Testing P7 Realism)
+
+- **Mock at the trait, not the struct:** `trait Store { fn get(&self, id: &str) -> Result<User, Error>; }` in production; `#[automock] trait Store` (via `mockall`) in test. The trait is the contract.
+- **`mockall` for generated mocks:** `#[automock] trait Repo {}` generates `MockRepo` with `expect_*` methods. Each expectation is per-test; no shared mock state (Testing P2 Independence).
+- **Mock the boundary, not the unit:** mock `Repo`, not `UserService` (the unit). Mocking the unit under test tests the mock (Testing P7 realism).
+- **No `#[cfg(test)]` on production code paths to inject mocks:** instead, accept the trait as a generic or `dyn` parameter. Test-only branches in production code are dead code in prod.
+
+```rust
+use mockall::*;
+
+#[automock]
+trait UserRepo {
+    fn get(&self, id: &str) -> Result<User, Error>;
+}
+
+#[test]
+fn get_user_returns_not_found() {
+    let mut repo = MockUserRepo::new();
+    repo.expect_get()
+        .with(eq("abc"))
+        .returning(|_| Err(Error::NotFound));
+    let svc = UserService::new(Box::new(repo));
+    assert!(matches!(svc.get_user("abc"), Err(Error::NotFound)));
+}
+```
+
+## Async Tests (Concurrency P10 Test for Race Conditions, Testing P1 Tests as Specification)
+
+- **`#[tokio::test]` for `async fn` tests:** runs the coroutine on a tokio runtime. Without it, an `async fn` test returns a future, never awaited (silently passes).
+- **`#[tokio::test(flavor = "multi_thread")]` for concurrency-sensitive tests:** multi-thread runtime surfaces races that single-thread misses (Concurrency P10).
+- **`tokio::time::pause()` and `advance()` for time:** freeze and advance the runtime clock deterministically. No `tokio::time::sleep(real)` in tests.
+- **Race-sensitive tests use `loom` for model-checking:** `loom` simulates all thread interleavings; it catches races `-race`-style detectors miss. Use for lock-free data structures.
+
+```rust
+#[tokio::test]
+async fn async_fetch_returns_user() {
+    let u = fetch_user("abc").await.unwrap();
+    assert!(!u.name.is_empty());
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
+async fn concurrent_cache_is_safe() {
+    let c = Arc::new(Cache::new());
+    let mut h = vec![];
+    for i in 0..10 {
+        let c = c.clone();
+        h.push(tokio::spawn(async move { c.get(&i.to_string()).await; }));
+    }
+    for x in h { x.await.unwrap(); }
+}
+```
+
+## Determinism and Time (Testing P3 Determinism, Testing P9 Edge Case Coverage)
+
+- **No `SystemTime::now()` or `Instant::now()` in code under test:** inject a `Clock` trait. In tests, a fake clock advances deterministically.
+- **`tokio::time::pause()` for async time:** freezes the runtime clock; `tokio::time::advance(dur)` moves it. A `sleep(5s)` in test resolves instantly.
+- **`--test-threads=1` to reproduce order coupling:** by default, `cargo test` runs tests in parallel; a test that passes alone but fails in a suite has hidden shared state. `-1` reproduces.
+
+```rust
+trait Clock { fn now(&self) -> std::time::Instant; }
+
+struct FakeClock(std::time::Instant);
+impl Clock for FakeClock {
+    fn now(&self) -> std::time::Instant { self.0 }
+}
+
+#[test]
+fn user_has_created_at() {
+    let clk = FakeClock(std::time::Instant::now());
+    let u = create_user_with_clock("a@b.co", &clk).unwrap();
+    assert_eq!(u.created_at, clk.now());
+}
+```
+
+## Doc Tests (Documentation P1 Documentation is Code, Testing P1 Tests as Specification)
+
+- **`cargo test --doc` runs `///` fenced blocks:** a `///` example with `#`-hidden setup is a tested artifact; a stale output fails the build (Documentation P1).
+- **`no_run` for examples that should compile but not run:** ```` ```rust,no_run ```` — type-checks the example without executing. Use for examples that need a DB.
+- **`ignore` for examples that should not compile-check:** ```` ```rust,ignore ```` — skips entirely. Rare; prefer `no_run`.
+- **Applies `Testing P1`:** the doc example is the spec; the doc test is the spec's regression test.
+
+```rust
+/// Fetch a user by id.
+///
+/// # Example
+///
+/// ```
+/// # use mycrate::{get_user, Error};
+/// let u = get_user("abc").unwrap();
+/// assert!(!u.name.is_empty());
+/// ```
+pub fn get_user(id: &str) -> Result<User, Error> { /* ... */ }
+```
+
+## Cross-References
+
+- `domains/testing/pyramid.md` — where unit/property/doc tests sit; proptest is the property layer.
+- `domains/testing/fixtures.md` — `t.Cleanup`-equivalent (`Drop` in tests) as fixture discipline.
+- `domains/testing/first-principles.md` — Testing P1 Specification, P2 Independence, P3 Determinism, P9 Edge Coverage.
+- `domains/concurrency/first-principles.md` — Concurrency P10 (test for races), `loom` model-checking.
+- `languages/rs-ownership.md` — `Send`/`Sync` tests and ownership-based property tests.
+- `languages/rs-async.md` — `#[tokio::test]` patterns from that doc.
+- `languages/rs-tooling.md` — `cargo test` flags (`--doc`, `--test-threads`) detailed here.
@@ -0,0 +1,95 @@
+# Rust Tooling — Derived Application
+
+> Applies Atelier's domain principles to Rust tooling specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/rust.md` for the language first-principles stub.
+
+## cargo and Build Discipline (DevOps P2 Automation, DevOps P1 Reproducibility)
+
+- **`cargo build` for dev, `cargo build --release` for release:** release enables optimizations (LTO, codegen-units=1). The default profile is for fast iteration, not perf.
+- **`Cargo.lock` committed for applications and CI:** for libraries, commit the lock for CI reproducibility even though consumers resolve their own tree. A drifted lock breaks reproducibility (DevOps P1).
+- **`cargo update` periodically, with a CI check:** `cargo update` bumps patch versions in the lock; a CI job that fails on lock drift catches a forgotten `cargo update`.
+- **`cargo vendor` for hermetic CI:** vendors `vendor/` into the repo; CI builds without network. The trade-off is repo size; the win is reproducibility.
+
+```toml
+# Cargo.toml — profile discipline
+[profile.release]
+lto = true
+codegen-units = 1
+panic = "abort"   # smaller binary, no unwinding
+```
+
+## clippy (DevOps P2 Automation, C2 Clarity)
+
+- **`cargo clippy` is the lint layer over `rustc`:** it catches `clone()` where a borrow would do, `unwrap()` in library code, and needless `Box`. Run on every build.
+- **`cargo clippy -- -D warnings` in CI:** warnings are errors. A clippy warning is a smell; accumulating them erodes the signal (Clarity C2).
+- **Per-lint allow only with a tracked reason:** `#[allow(clippy::needless_collect)] // reason: GH-123 — collect needed for len` — each allow links to a ticket. Untracked allows accumulate into a permanently lint-bypassed core.
+- **`cargo clippy --fix` for safe auto-fixes:** applies the linter's suggested change. Review the diff; do not run blindly on a large commit.
+
+```bash
+# CI gate
+cargo clippy --all-targets --all-features -- -D warnings
+```
+
+## cargo fmt (DevOps P2 Automation, C2 Clarity)
+
+- **`cargo fmt` is the formatter; format is not debated in review:** run in CI as a check (`cargo fmt --check`), not a fix. A failing check blocks the PR.
+- **`rustfmt.toml` for repo-wide settings:** if the defaults are wrong for the repo, override once and stop. Do not relitigate per-PR.
+- **Applies `devops/P2`:** the format gate is automated; a reviewer never comments on style.
+
+```bash
+# CI gate — fail if unformatted
+cargo fmt --check
+```
+
+## Edition Discipline (DevOps P1 Reproducibility, C5 Reversibility)
+
+- **`edition` in `Cargo.toml` pins the language edition:** 2015, 2018, 2021, 2024. An edition is a coherent set of language changes; bumping it is a deliberate migration.
+- **Edition is not the compiler version:** `rustc 1.75` supports edition 2021; edition 2024 needs a newer `rustc`. Pin the toolchain with `rust-toolchain.toml`.
+- **Bump editions deliberately, not opportunistically:** `cargo fix --edition` applies the migration lint; review the diff. A bump mid-feature conflates two changes.
+- **Applies `devops/P1` and `C5` (reversibility):** pinning the edition and toolchain makes the build reproducible; bumping is a controlled, reversible change.
+
+```toml
+# Cargo.toml
+[package]
+edition = "2021"
+rust-version = "1.75"
+```
+
+```toml
+# rust-toolchain.toml
+[toolchain]
+channel = "1.75"
+components = ["clippy", "rustfmt"]
+```
+
+## Documentation in the Pipeline (Documentation P1 Documentation is Code, DevOps P9 Documentation in the Pipeline)
+
+- **`cargo doc` from doc comments:** `///` on items generates API docs; `cargo doc --open` previews. The build fails on broken intra-doc links (`#![warn(rustdoc::broken_intra_doc_links)]`).
+- **Doc tests are run by `cargo test`:** a `///` fenced block with `#`-hidden setup is a tested artifact; a stale example fails `cargo test --doc` (Documentation P1).
+- **`#![warn(missing_docs)]` for libraries:** public items without doc comments fail the build. Documentation is a build gate, not an afterthought.
+- **`cargo readme` or `cargo docs-rs` for landing pages:** the crate's `README.md` is rendered on docs.rs; keep it in sync with `lib.rs`'s top-level doc.
+
+```rust
+#![warn(missing_docs, rustdoc::broken_intra_doc_links)]
+
+/// Fetch a user by id.
+///
+/// # Example
+///
+/// ```
+/// # use mycrate::get_user;
+/// let u = get_user("abc").unwrap();
+/// println!("{}", u.name);
+/// ```
+pub fn get_user(id: &str) -> Result<User, Error> { /* ... */ }
+```
+
+## Cross-References
+
+- `domains/devops/ci-cd.md` — the pipeline gates that host clippy/fmt/test.
+- `domains/devops/first-principles.md` — DevOps P1 Reproducibility, P2 Automation.
+- `domains/documentation/first-principles.md` — Documentation P1 Documentation is Code.
+- `languages/rs-ownership.md` — `Send`/`Sync` clippy lints reference this doc.
+- `languages/rs-async.md` — async-runtime tooling (`tokio` features) detailed here.
+- `languages/rs-testing.md` — `cargo test` flags (`--doc`, `--no-run`) detailed here.
@@ -2,6 +2,13 @@

 > How Atelier's domain principles apply in Rust specifically. Derives from `domains/` docs.

+## Derived Docs
+
+- [rs-ownership.md](rs-ownership.md) — Send/Sync, lifetimes, borrowing, ownership transfer.
+- [rs-tooling.md](rs-tooling.md) — cargo, clippy, fmt, edition discipline.
+- [rs-async.md](rs-async.md) — tokio, async traits, cancellation, pin.
+- [rs-testing.md](rs-testing.md) — #[test], proptest, property testing, mock discipline.
+
 ## Type System (C1 Correctness, Data P7 Type Fidelity)

 - **Newtypes for domain concepts:** `struct UserId(String);` — zero-cost, type-safe.
@@ -0,0 +1,115 @@
+# TypeScript Async — Derived Application
+
+> Applies Atelier's domain principles to TypeScript async specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/typescript.md` for the language first-principles stub.
+
+## Promises and AbortSignal (Concurrency P7 Cancellation Support, C1 Correctness)
+
+- **Every async function accepts an optional `AbortSignal`:** cancellation is a first-class parameter, not a side channel. The signal propagates to `fetch`, `setTimeout`, and downstream awaits.
+- **`AbortController` is the producer side; `AbortSignal` is the consumer side:** a function takes a `signal` (read-only), the caller owns the `controller` and decides when to abort.
+- **Abort propagates as a rejected `Promise`:** `fetch` rejects with `AbortError`; downstream code sees the rejection, not a silent no-op. This preserves `errors/P5` (recoverable when possible) — the caller can distinguish cancellation from a real failure.
+- **Applies `concurrency/P7`:** no async operation runs without a path to cancel it. A long-running `await` with no signal is a hung request.
+- **Never swallow `AbortError`:** re-throw or handle distinctly; cancellation is the caller's intent, not an error to log.
+
+```typescript
+async function fetchUser(id: UserId, signal?: AbortSignal): Promise<User> {
+  const ctrl = new AbortController();
+  signal?.addEventListener('abort', () => ctrl.abort());
+  const res = await fetch(`/users/${id}`, { signal: ctrl.signal });
+  if (!res.ok) throw new HttpError(res.status);
+  return res.json() as Promise<User>;
+}
+
+// caller controls cancellation
+const ctrl = new AbortController();
+const timer = setTimeout(() => ctrl.abort(), 5000);
+try {
+  const u = await fetchUser(id, ctrl.signal);
+} finally {
+  clearTimeout(timer);
+}
+```
+
+## async/await Discipline (Concurrency P8 Timeout Discipline, C2 Clarity)
+
+- **`await` is the only async primitive in application code:** no `.then` chains, no callback pyramids. `async`/`await` reads top-to-bottom (Clarity C2).
+- **Never `await` in a hot loop without batching:** sequential `await` in a `for` loop is O(n) latency. Use `Promise.all` for parallelism; `for await...of` only for genuine streams.
+- **`Promise.race` for a timeout:** every external `await` has a deadline. `Promise.race([op, timeout])` rejects when the deadline passes.
+- **`return` vs `return await`:** inside `try`/`finally`, `return await` runs the `finally`; bare `return` of a Promise defers the `finally` to the microtask. Prefer `return await` when cleanup must run.
+- **Applies `concurrency/P8`:** a bare `await` with no timeout is an unbounded wait. External calls (network, disk) always race against a deadline.
+
+```typescript
+async function fetchWithTimeout(url: string, ms = 5000, signal?: AbortSignal): Promise<Response> {
+  const ctrl = new AbortController();
+  signal?.addEventListener('abort', () => ctrl.abort());
+  const timer = new Promise<never>((_, reject) =>
+    setTimeout(() => reject(new TimeoutError(ms)), ms)
+  );
+  try {
+    return await Promise.race([fetch(url, { signal: ctrl.signal }), timer]);
+  } finally {
+    clearTimeout(timer); // cleanup runs on success and on race-loss
+  }
+}
+```
+
+## Error Handling in Async (Errors P5 Recoverable When Possible, Errors P1 Errors are Data)
+
+- **Catch `unknown`, narrow with a type guard:** `catch (e: unknown)` — TS does not infer the error type. `instanceof` or a discriminator narrows it.
+- **Retry with backoff for transient failures:** network blips are recoverable (Errors P5). Exponential backoff with jitter, capped retry count, and an `AbortSignal`-aware `setTimeout`.
+- **No retry for non-idempotent operations:** a `POST` that creates a resource is not safely retryable without an idempotency key (applies `api/P6` Idempotency).
+- **Typed errors over `Error` subclasses:** a discriminated union `AppError = Network | Timeout | Cancelled` carries context (Errors P4 Preserve Context) without `instanceof` chains.
+
+```typescript
+async function fetchRetry(url: string, attempts = 3, signal?: AbortSignal): Promise<Response> {
+  for (let i = 0; i < attempts; i++) {
+    try {
+      return await fetchWithTimeout(url, 5000, signal);
+    } catch (e: unknown) {
+      if (e instanceof AbortError) throw e;          // do not retry cancellation
+      if (e instanceof TimeoutError && i < attempts - 1) {
+        await sleep(jitter(i), signal);             // backoff before retry
+        continue;
+      }
+      throw e;
+    }
+  }
+  throw new Error('unreachable');
+}
+```
+
+## Cancellation Propagation (Concurrency P7 Cancellation Support, Concurrency P9 Bounded Queues)
+
+- **One signal, many consumers:** pass the same `AbortSignal` to every async call in a request. Aborting once cancels the whole tree.
+- **Bounded concurrency with a semaphore:** a `Semaphore(N)` wrapping `Promise.all` caps in-flight requests (Concurrency P9 — bounded queues). Unbounded `Promise.all` on a 10k-item array exhausts file descriptors.
+- **Cancellation is cooperative, not preemptive:** a long synchronous block inside an `async` function ignores the signal. Yield with `await Promise.resolve()` periodically in CPU-bound loops, or move to a worker.
+- **Applies `messaging/delivery-semantics`:** a cancelable async operation is an at-most-once delivery — the caller may stop listening, the result may or may not arrive. Retry-on-cancel is at-least-once; the caller must declare which.
+
+```typescript
+async function mapBounded<T, U>(items: readonly T[], fn: (t: T, s: AbortSignal) => Promise<U>, limit = 8, signal?: AbortSignal): Promise<U[]> {
+  const ctrl = new AbortController();
+  signal?.addEventListener('abort', () => ctrl.abort());
+  const results: U[] = new Array(items.length);
+  let next = 0;
+  const workers = Array.from({ length: limit }, async () => {
+    while (true) {
+      const i = next++;
+      if (i >= items.length) break;
+      if (ctrl.signal.aborted) throw new AbortError();
+      results[i] = await fn(items[i], ctrl.signal);
+    }
+  });
+  await Promise.all(workers);
+  return results;
+}
+```
+
+## Cross-References
+
+- `domains/concurrency/patterns.md` — the cancellation/timeout/semaphore patterns applied here.
+- `domains/concurrency/first-principles.md` — Concurrency P7 Cancellation Support, P8 Timeout Discipline, P9 Bounded Queues.
+- `domains/messaging/delivery-semantics.md` — at-most-once vs at-least-once framing for async retry/cancel (IDEATE-40).
+- `domains/errors/patterns.md` — typed async errors and retry-with-backoff.
+- `languages/ts-types.md` — `Result<T, E>` and discriminated `AppError` used in async error handling.
+- `languages/ts-tooling.md` — `no-floating-promises` lint rule that enforces these awaits.
@@ -0,0 +1,117 @@
+# TypeScript Testing — Derived Application
+
+> Applies Atelier's domain principles to TypeScript testing specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/typescript.md` for the language first-principles stub.
+
+## Vitest and Jest (Testing P1 Tests as Specification, C2 Clarity)
+
+- **Vitest for new TS projects; Jest for legacy:** Vitest shares `vite`'s transform pipeline (no separate `ts-jest` config); Jest's ecosystem is broader. Either is acceptable — pick one per repo, do not mix.
+- **Tests co-located with source:** `user.ts` → `user.test.ts`. A test file far from its subject rots (Documentation P5 Discoverability).
+- **`describe`/`it` mirror the public API:** the test block names read as a specification ("User", "rejects an invalid email", "returns the persisted id"). A reader should understand the unit from test names alone (Testing P1).
+- **`expect` over `assert`:** Vitest/Jest matchers produce readable failure output (`expect(x).toBe(y)` → "expected 5, received 3"). Raw `assert` gives a stack trace and nothing else (Testing P6 Failure Specificity).
+
+```typescript
+// user.test.ts
+import { describe, it, expect } from 'vitest';
+import { createUser } from './user';
+
+describe('createUser', () => {
+  it('rejects an invalid email', async () => {
+    await expect(createUser({ email: 'not-an-email' })).rejects.toThrow(ValidationError);
+  });
+  it('returns the persisted id', async () => {
+    const u = await createUser({ email: 'a@b.co' });
+    expect(u.id).toMatch(/^[a-z0-9]+$/);
+  });
+});
+```
+
+## Mock Discipline (Testing P2 Independence, Testing P7 Realism)
+
+- **Mock at the boundary, not the unit:** replace `fetch` or the DB client, not the function under test. Mocking the unit under test tests the mock, not the code (Testing P7 — realism).
+- **No partial mocks of the system under test:** if a method must be stubbed, the unit is too large. Extract a collaborator and mock that.
+- **Each test sets up and tears down its own state:** no shared mutable fixtures. A `beforeEach`/`afterEach` resets; a top-level `let` shared across tests is order-coupling (Testing P2 Independence).
+- **`vi.useFakeTimers()` for time-dependent code:** never call `Date.now()` directly in code under test; inject a `Clock` port. In tests, fake timers make `setTimeout` synchronous.
+
+```typescript
+import { vi, beforeEach, afterEach } from 'vitest';
+
+beforeEach(() => {
+  vi.useFakeTimers();
+  global.fetch = vi.fn();   // boundary mock
+});
+afterEach(() => {
+  vi.useRealTimers();
+  vi.restoreAllMocks();
+});
+```
+
+## Type-Level Tests (Testing P1 Tests as Specification, Data P7 Type Fidelity)
+
+- **Type-level tests assert the type system, not runtime behavior:** `expectTypeOf<T>().toMatchTypeOf<U>` and `tsd`/`expect-type` fail the build when a type assertion is wrong.
+- **Negative type tests are required:** `// @ts-expect-error` proves the compiler rejects what it should. A `@ts-expect-error` that no longer errors is itself an error (the comment must be consumed).
+- **Branded types and utility types get type tests:** a `UserId` should not be assignable to `string`; a `Readonly<T>` should not allow assignment. These invariants are part of the spec (Testing P1).
+- **Applies `data/P7` (type fidelity):** a type-level test is a regression test for the type checker — if a refactor silently widens a type, the test fails.
+
+```typescript
+import { expectTypeOf } from 'expect-type';
+import type { User, UserPatch, UserId } from './user';
+
+test('UserPatch omits id and makes fields optional', () => {
+  expectTypeOf<UserPatch>().toMatchTypeOf<{ name?: string; email?: string }>();
+  expectTypeOf<UserPatch>().not.toHaveProperty('id');
+});
+
+test('UserId is not assignable to bare string', () => {
+  // @ts-expect-error — brand prevents widening
+  const s: string = {} as UserId;
+  expect(s).toBeDefined();
+});
+```
+
+## Parametrize and Factories (Testing P3 Determinism, Testing P9 Edge Case Coverage)
+
+- **`it.each` / `test.each` for parametrized cases:** one table drives many runs; each row is an independent test with its own name and failure output.
+- **Factories over fixtures:** `makeUser(overrides)` returns a fresh object per call. A shared `const user = {...}` across tests couples them and breaks determinism when one test mutates it (Testing P3).
+- **Edge cases as rows, not special tests:** empty array, single element, max int, null, undefined — each a row in a `test.each` table. An ad-hoc `it('handles edge')` with multiple asserts hides which case failed (Testing P9 — edge case coverage, P6 failure specificity).
+- **Property-style tests via `fast-check`:** for invariants (e.g., "parse(serialize(x)) === x"), `fast-check` generates hundreds of inputs and shrinks failures to a minimal counterexample.
+
+```typescript
+import { test, expect } from 'vitest';
+import { makeUser } from './user.factory';
+
+test.each([
+  { input: '', reason: 'empty' },
+  { input: 'a'.repeat(1000), reason: 'too long' },
+  { input: 'not-an-email', reason: 'no @' },
+])('rejects email: $reason', async ({ input }) => {
+  await expect(makeUser({ email: input })).rejects.toThrow(ValidationError);
+});
+```
+
+## Determinism and Time (Testing P3 Determinism, Testing P9 Edge Case Coverage)
+
+- **No `Date.now()`, `Math.random()`, or `crypto.randomUUID()` in code under test:** inject a `Clock`, `Random`, and `IdGen` port. In tests, provide deterministic fakes.
+- **`--random` test order (Vitest `sequence.shuffle: true` default) catches order coupling:** a test that passes alone but fails in a suite has hidden state. The shuffle makes that state visible (Testing P2).
+- **Race-detector parallelism for async tests:** run async tests concurrently by default; a test that assumes serial execution breaks under parallelism. Vitest's `concurrent` flag surfaces the bug.
+
+```typescript
+import { vi, test, expect } from 'vitest';
+
+test.concurrent('parallel fetch does not interleave state', async () => {
+  const store = new Store();
+  await Promise.all([store.put('a', 1), store.put('b', 2)]);
+  expect(store.get('a')).toBe(1);
+  expect(store.get('b')).toBe(2);
+});
+```
+
+## Cross-References
+
+- `domains/testing/pyramid.md` — where unit/type/integration tests sit; the type-level tests here are the base layer.
+- `domains/testing/fixtures.md` — factory-vs-fixture discipline applied via `makeUser`.
+- `domains/testing/first-principles.md` — Testing P1 Specification, P2 Independence, P3 Determinism, P9 Edge Coverage.
+- `languages/ts-types.md` — the branded types and utility types that type-level tests assert.
+- `languages/ts-async.md` — async tests use the cancellation/timeout patterns from that doc.
+- `languages/ts-tooling.md` — `ts-jest`/`vitest` config and the `expect-type`/`tsd` toolchain.
@@ -0,0 +1,114 @@
+# TypeScript Tooling — Derived Application
+
+> Applies Atelier's domain principles to TypeScript tooling specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/typescript.md` for the language first-principles stub.
+
+## tsc and tsconfig Discipline (DevOps P2 Automation, DevOps P1 Reproducibility)
+
+- **`strict: true` is the floor, not the ceiling:** it enables `strictNullChecks`, `noImplicitAny`, `strictFunctionTypes`, and more. Disable sub-flags only with a justification comment.
+- **`tsc --noEmit` in CI:** type-checking is a build gate; emission is the bundler's job. Separate the two so a type error fails CI even when the bundler would have succeeded.
+- **`tsconfig` is per-project, not inherited verbatim:** a shared base (`extends`) encodes org defaults; each project overrides the deltas it needs. Avoids the "one monoreto-config-fits-all" trap.
+- **`noUncheckedIndexedAccess` for safety:** `arr[i]` becomes `T | undefined`, forcing narrowing. Costs little, prevents a class of out-of-bounds deref bugs.
+- **Applies `devops/P1` (reproducibility):** pinned `typescript` version in `package.json` and `lockfile` ensure every CI run type-checks against the same compiler.
+
+```jsonc
+// tsconfig.json — base
+{
+  "compilerOptions": {
+    "strict": true,
+    "noUncheckedIndexedAccess": true,
+    "exactOptionalPropertyTypes": true,
+    "noEmit": true,
+    "moduleResolution": "bundler",
+    "isolatedModules": true
+  }
+}
+```
+
+## ESLint and @typescript-eslint (DevOps P2 Automation, Documentation P9 Living Documents)
+
+- **ESLint with `@typescript-eslint` strict ruleset:** `recommended-type-checked` enables rules that require the type checker (`no-floating-promises`, `no-misused-promises`).
+- **Rules encode decisions, not taste:** every custom rule in the config has a one-line `// reason:` comment linking to the principle it enforces. This makes the config a living document (Documentation P9).
+- **Format is Prettier's job; ESLint lints:** `eslint-config-prettier` disables conflicting format rules. Do not relitigate formatting in code review.
+- **`no-floating-promises` enforces `concurrency/P8` (timeout discipline):** an un-awaited `Promise` is a fire-and-forget that swallows errors and timeouts. The rule forces `.catch()` or `await`.
+
+```jsonc
+// .eslintrc.json
+{
+  "extends": [
+    "eslint:recommended",
+    "plugin:@typescript-eslint/recommended-type-checked",
+    "prettier"
+  ],
+  "parserOptions": { "project": "./tsconfig.json" },
+  "rules": {
+    // reason: enforce Concurrency P8 — no un-awaited promises
+    "@typescript-eslint/no-floating-promises": "error",
+    // reason: enforce Data P7 — no `any` escaping the type checker
+    "@typescript-eslint/no-explicit-any": "error"
+  }
+}
+```
+
+## Project References and ts-jest (DevOps P2 Automation, C6 Composability)
+
+- **Project references for monorepos:** `composite: true` + `references` let `tsc --build` incrementally type-check only changed projects, and enforce the dependency graph at the type level.
+- **`paths` aliases mirror the import structure:** `@app/*` → `src/*`. Configure once in `tsconfig`, mirror in the bundler and the test runner so all three agree.
+- **`ts-jest` (or `vitest`) with `isolatedModules: true`:** each test file is type-checked in isolation, matching how the bundler transpiles. Catches the "passes in `tsc` but fails in the bundler" gap.
+- **Applies `devops/P2`:** the build pipeline (tsc → lint → test → bundle) is automated; a developer never runs a manual sequence.
+
+```jsonc
+// tsconfig.references.json
+{
+  "files": [],
+  "references": [
+    { "path": "./packages/core" },
+    { "path": "./packages/api" },
+    { "path": "./packages/web" }
+  ]
+}
+```
+
+## Lockfile and Reproducible Install (DevOps P1 Reproducibility)
+
+- **`npm ci` in CI, not `npm install`:** `ci` reads the lockfile exactly and fails on drift. `install` mutates the lockfile.
+- **Lockfile committed for applications:** for libraries, commit `package-lock.json` for CI reproducibility even though consumers resolve their own tree.
+- **No floating ranges in `package.json`:** `^` and `~` are CI's job to resolve; pin the resolved version in the lockfile. An unpinned `*` is a supply-chain attack surface.
+
+```bash
+# CI install step — deterministic
+npm ci
+# Type-check gate
+npx tsc --noEmit
+# Lint gate
+npx eslint .
+```
+
+## Documentation in the Pipeline (Documentation P1 Documentation is Code, DevOps P9 Documentation in the Pipeline)
+
+- **Type-checked JSDoc:** `typedoc` (or `TypeDoc`) generates API docs from `tsdoc` comments. The compiler enforces that `@param` names match real parameters.
+- **`@example` blocks are compiled:** a `tsdoc` `@example` fenced block is type-checked as part of the doc build. Stale examples fail the pipeline (Documentation P1 — docs are code).
+- **README badges reflect CI status:** the build/lint/test/type-check gates are the source of truth; badges surface them. Do not hand-edit status tables.
+
+```typescript
+/**
+ * Fetch a user by ID.
+ *
+ * @param id - a branded UserId (see ts-types.md).
+ * @throws {NotFoundError} if the user does not exist.
+ * @example
+ * ```ts
+ * const u = await getUser(userId('abc'));
+ * ```
+ */
+async function getUser(id: UserId): Promise<User> { /* ... */ }
+```
+
+## Cross-References
+
+- `domains/devops/ci-cd.md` — the pipeline gates that host tsc/ESLint/ts-jest.
+- `domains/devops/first-principles.md` — DevOps P1 Reproducibility, P2 Automation.
+- `domains/documentation/first-principles.md` — Documentation P1 Documentation is Code.
+- `languages/ts-types.md` — the type rules ESLint enforces reference this doc.
+- `languages/ts-testing.md` — the test-runner config (`ts-jest`/`vitest`) detailed here.
@@ -0,0 +1,111 @@
+# TypeScript Type System — Derived Application
+
+> Applies Atelier's domain principles to TypeScript's type system specifically.
+> Derives from `domains/` docs; introduces no new P-rules (D-063).
+> See `languages/typescript.md` for the language first-principles stub.
+
+## Nominal vs Structural Typing (C1 Correctness, Data P7 Type Fidelity, API P1 Contract Fidelity)
+
+- **TypeScript is structurally typed:** two types with the same shape are assignable. This is convenient but erases domain boundaries — a `UserId` and `PostId` both `string` are interchangeable.
+- **Branded (nominal) types for domain IDs:** intersect with a phantom brand to simulate nominal typing. The brand is never constructed at runtime; it exists only to the type checker.
+- **Applies `data/P7` (type fidelity)** at the value boundary: a branded `UserId` cannot be passed where a `PostId` is expected, preventing an entire class of swap bugs.
+- **Applies `api/P1` (contract fidelity):** branded types make API contracts explicit — handlers cannot accept "any string" for an ID.
+- **Brand is opaque to consumers:** do not export the brand symbol; construction goes through a validated factory.
+
+```typescript
+type UserId = string & { readonly __brand: 'UserId' };
+type PostId = string & { readonly __brand: 'PostId' };
+
+function userId(s: string): UserId {
+  if (!/^[a-zA-Z0-9]+$/.test(s)) throw new Error('invalid id');
+  return s as UserId;
+}
+
+function getUser(id: UserId): User { /* ... */ }
+getUser('abc');            // type error
+getUser(userId('abc'));   // ok
+getUser(postId('xyz'));   // type error — distinct brands
+```
+
+## Generics (C6 Composability, Data P7 Type Fidelity)
+
+- **Generics preserve type information across boundaries:** a `Repository<T>` keeps the element type through `find`/`save` rather than widening to `any`.
+- **Constrain with `extends`:** `<T extends Entity>` documents the contract and gives the body access to `T.id`.
+- **Avoid unnecessary generics:** if a function accepts "any value and returns it unchanged," `T` is noise. Prefer `unknown` for truly opaque inputs.
+- **Variance is structural:** TS does not enforce sound variance; mark mutation points with `readonly` to keep `T[]` assignable to `readonly T[]`.
+
+```typescript
+interface Entity { id: string }
+class Repository<T extends Entity> {
+  constructor(private db: Map<string, T>) {}
+  find(id: string): T | undefined { return this.db.get(id); }
+  save(t: T): void { this.db.set(t.id, t); }
+}
+```
+
+## Narrowing and Type Guards (C1 Correctness, Errors P1 Errors are Data)
+
+- **Narrowing is how TS handles `unknown` and union types safely:** `typeof`, `in`, `instanceof`, and discriminators collapse a wide type to a precise one before use.
+- **User-defined type guards (`x is T`) encode domain predicates:** `isUser(x): x is User` lets the checker track the narrow across call sites.
+- **Applies `errors/P1` (errors are data):** a `Result<T, E>` discriminated union is narrowed with `if (r.ok)` — no `try`/`catch` needed for expected failures.
+- **Never use `as` to widen past a check:** `as` lies to the compiler. If narrowing does not reach the type you need, the predicate is wrong, not the cast.
+
+```typescript
+type Result<T, E> = { ok: true; value: T } | { ok: false; error: E };
+
+function unwrap<T, E>(r: Result<T, E>): T {
+  if (r.ok) return r.value;       // narrowed to { ok: true; value: T }
+  throw new Error(JSON.stringify(r.error));
+}
+
+function isUser(x: unknown): x is User {
+  return typeof x === 'object' && x !== null && 'id' in x && 'name' in x;
+}
+```
+
+## Utility Types (C5 Reversibility, C6 Composability)
+
+- **`Partial<T>`, `Pick<T,K>`, `Omit<T,K>`, `Readonly<T>` are derived views:** they derive from a source-of-truth `T` rather than redeclaring fields, so the source change propagates (reversibility).
+- **`Readonly<T>` enforces immutability at the type level** — applies `concurrency/P1` (immutability by default) without runtime cost.
+- **`Record<K, V>` over `{ [k: string]: V }`:** the index signature form allows any string key including prototype pollution vectors; `Record` is exact.
+- **Compose, don't accumulate:** `type Patch<T> = Partial<Omit<T, 'id'>>` reads as a transformation; restate it if `T` changes shape, rather than maintaining a parallel `Patch` type.
+
+```typescript
+interface User { id: string; name: string; email: string; }
+type UserPatch = Partial<Omit<User, 'id'>>;
+type ReadonlyUser = Readonly<User>;
+type UsersById = Record<string, User>;
+```
+
+## Discriminated Unions (C1 Correctness, Data P7 Type Fidelity, Errors P1 Errors are Data)
+
+- **Discriminated unions over enums:** `type Status = { type: 'pending' } | { type: 'paid'; amount: number }` is exhaustive and carries payload per variant; an `enum` carries neither.
+- **The discriminant is a literal `type` (or `kind`) field:** the checker narrows on it in `switch` and `if` without a custom guard.
+- **Exhaustiveness via `never`:** assign the narrowed value to `never` in the default branch; if a variant is added, the default fails to compile.
+- **Applies `errors/P1`:** model domain errors as a discriminated union `AppError = NotFound | Validation | Conflict`, not as exception classes — the type system carries the error set.
+
+```typescript
+type Status =
+  | { type: 'pending' }
+  | { type: 'paid'; amount: number }
+  | { type: 'refunded'; reason: string };
+
+function describe(s: Status): string {
+  switch (s.type) {
+    case 'pending':  return 'awaiting payment';
+    case 'paid':     return `paid ${s.amount}`;
+    case 'refunded': return `refunded: ${s.reason}`;
+    default:
+      const _exhaustive: never = s;  // compile error if a variant is added
+      throw new Error('unhandled');
+  }
+}
+```
+
+## Cross-References
+
+- `domains/data/schema-design.md` — schema-level fidelity parallels branded types at the TS boundary.
+- `domains/data/first-principles.md` — Data P7 Type Fidelity, the primary trace for this doc.
+- `domains/api/rest.md` — contract fidelity for API handlers consuming branded IDs.
+- `domains/errors/patterns.md` — discriminated unions as the error-as-data encoding.
+- `languages/ts-async.md` — typed async results built on the `Result` union here.
@@ -2,6 +2,13 @@

 > How Atelier's domain principles apply in TypeScript specifically. Derives from `domains/` docs; this file is the language-specific lens.

+## Derived Docs
+
+- [ts-types.md](ts-types.md) — TS type system: nominal-via-branding, generics, narrowing, utility types, discriminated unions.
+- [ts-tooling.md](ts-tooling.md) — tsc, ESLint, ts-jest, project references, tsconfig discipline.
+- [ts-async.md](ts-async.md) — Promises + AbortSignal, async/await, error handling, cancellation.
+- [ts-testing.md](ts-testing.md) — Vitest/Jest, mock discipline, type-level tests.
+
 ## Type System (C1 Correctness, Data P7 Type Fidelity)

 - **Strict mode on:** `strict: true` in `tsconfig.json`. No `any` without justification.
Author	SHA1	Message	Date
Jon Chery	4e433158cd	docs(P03): complete language-derived extension — v0.4 ---ci--- project: atelier phase: 3 milestone: v0.4 status: complete phase_role: execution phase_tag: v0.3.3 requirements: covered: [ATELIER-102, ATELIER-103, ATELIER-104, ATELIER-105] partial: [] ---/ci---	2026-08-05 16:07:21 +00:00
Jon Chery	0fdf892d81	docs(ship): P2 complete — v0.3.2 tagged, release 483 ---ci--- project: atelier phase: 2 milestone: v0.4 status: complete phase_role: execution phase_tag: v0.3.2 release_id: 483 ---/ci---	2026-08-05 16:01:08 +00:00
Jon Chery	f61dffbb5a	docs(P02): complete messaging domain phase — v0.4 ---ci--- project: atelier phase: 2 milestone: v0.4 status: complete phase_role: execution phase_tag: v0.3.2 requirements: covered: [ATELIER-97, ATELIER-98, ATELIER-99, ATELIER-100, ATELIER-101] partial: [] ---/ci---	2026-08-05 16:00:31 +00:00
Jon Chery	f7f007dce8	docs(ship): P1 complete — v0.3.1 tagged, release 479 ---ci--- project: atelier phase: 1 milestone: v0.4 status: complete phase_role: execution phase_tag: v0.3.1 release_id: 479 ---/ci---	2026-08-05 15:53:16 +00:00
Jon Chery	20992883ff	docs(P01): complete edge domain phase — v0.4 ---ci--- project: atelier phase: 1 milestone: v0.4 status: complete phase_role: execution phase_tag: v0.3.1 requirements: covered: [ATELIER-92, ATELIER-93, ATELIER-94, ATELIER-95, ATELIER-96] partial: [] ---/ci---	2026-08-05 15:51:03 +00:00
Jon Chery	a622419e7d	docs(ship): P0 complete — v0.3.0 tagged, release 475 ---ci--- project: atelier phase: 0 milestone: v0.4 status: complete phase_role: pre_execution phase_tag: v0.3.0 release_id: 475 ---/ci---	2026-08-05 14:03:35 +00:00