문제: registry의 CACHE_INVALIDATION_PULSE 행은 backend localStorage,
valueCodec opaque-string-v1이다. 실제로 이 키에 쓰는 쪽은 cross-context
어댑터이고 envelope 없는 wire JSON을 setItem으로 바로 쓴다. 그래서
StoragePort로 read하면 Web Storage 어댑터가 값을 손상 envelope로 보고
물리 키를 지우고, write하면 envelope 문자열로 덮어써 수신 탭이 버린다.
고른 방법: StorageBackend에 "external"을 더하고 pulse 행을
backend external, valueCodec none으로 바꾼다. getStorageDefinition이
external 키를 던지므로 어댑터의 read/write/remove는 물리 키를 건드리지
않고 *_UNAVAILABLE 실패를 돌려준다. defineStorageKey는 external 키에
codec이 붙으면 거절해 같은 거짓 행이 다시 생기지 않게 한다. 물리 키
문자열 ca-frontend:cache-invalidation:v1:pulse와 wire 형식은 그대로다.
registry-governance의 backend 허용값에 external을 더하고, breaking
필드 변경 두 건은 registry-change-evidence에 근거를 남긴다(baseline
재승인 대신 기존 08-13·F-6과 같은 evidence 방식).
버린 대안:
- 행 위 주석만: 가장 싸지만 StoragePort 호출을 막지 못한다.
- cross-context 어댑터가 StoragePort로 쓰게 함: storage 이벤트의
newValue 형식이 바뀌어 이미 배포된 탭과 서로 못 읽는다.
위험·한계: StoragePort 공개 시그니처는 그대로다. StorageBackend
유니온에 값이 하나 늘어(추가만) exhaustive switch를 쓰는 다운스트림은
분기를 더해야 한다. check:registries 결과는 compatibility=breaking
(evidence로 통과)다. 기본 구성에는 이 키를 StoragePort로 쓰는 호출이
없어 실제 동작 변화는 없다.
검증:
- RED: storage-registry "refuses the externally owned pulse key..."가
"expected { ok: true, value: undefined } to match object { ok: false }"
로 실패(손상 discard 경로), codec 거절 테스트는 "expected [Function]
to throw an error"로 실패.
- GREEN: 같은 파일 23/23 통과.
- 회귀: storage-registry, registry-governance, cross-tab-invalidation,
contract-registry-lookup, browser-security, runtime-adapters,
release-mismatch-recovery 7파일 87/87 통과. tsc app·test 통과,
eslint 통과, check-registries PASS, verify-documentation-readiness
PASS_SCOPED. check:architecture는 pnpm 실행을 요구해 미실행.
Refs: docs/reviews/2026-11-01-maturity-review/05-adapters-storage-kernel.md
문제
초기 JS는 203,572 / 204,800 B(gzip)였다. 생성 인덱스가 각 기능의
feature-module.ts를 정적으로 import하고, 그 파일이 계약(zod 스키마·mapper
포함), HTTP 계약 contribution, application input 팩토리, route codec까지
모두 끌고 왔다. 참조 모양 기능 1개를 더하면 +3,922 B가 늘어 첫 기능에서
예산을 넘었다.
고른 방법
- 기능을 두 부분으로 나눈다. 선언(FeatureModule: route 표, route runtime
계약, invalidation, 문구, 부팅 때 release manifest와 맞춰 볼
contractPackages)은 초기 청크에 남는다. 실행부(FeatureRuntime: V2 API
계약, adapter contribution, contract contribution)는 feature-runtime.ts로
옮기고 loadRuntime이 동적 import한다. route codec은 context-routes.ts의
route handle에 실어 route와 함께 적재한다.
- 적재 시점은 그 기능의 route에 들어갈 때다. ContextRouteGateway가 route
모듈과 함께 application.features.load(featureId)를 부르고, 끝난 뒤에
화면을 그린다(공개 포트에는 load만 추가). 들어온 URL의 기능은 route
모듈과 병렬로 적재해 왕복을 늘리지 않는다. 런타임 설정으로 꺼진 기능은
적재하지 않는다.
- 조립부는 기능별 로더를 만든다(installed-feature-adapters.ts). contract
registry는 기능이 적재될 때마다 전체를 다시 합쳐 검증한다
(installed-contract-contributions.ts). 선언하지 않은 contract package를
적재하면 거절한다.
- 모든 실행부를 한꺼번에 읽는 registry(V2 API, contract contribution)는
installed-feature-runtime-registries.ts의 async 함수로, Node·테스트·
registry 거버넌스용 상수는 top-level await 모듈 loaded-feature-runtimes.ts로
둔다. src에서 후자를 import하면 새 아키텍처 규칙
boot-does-not-load-every-feature-runtime이 막는다. 플랫폼 구성 화면은
route 로더에서 async 함수를 불러 결과를 prop으로 받는다.
- vite.config.ts에 청크 그룹 2개를 둔다. rolldown은 부팅 그래프와 lazy
청크가 같이 쓰는 모듈을 importer 집합마다 따로 청크로 뗀다. 그대로 두면
기능마다 선언 청크가 하나씩 늘고(기능당 +1,471 B), 실행부와 공유하는
플랫폼 모듈(HTTP executor, contract runtime, zod)이 6개 청크로 쪼개져
템플릿 자체가 +438 B 커졌다. feature-declarations는 모든 기능의 선언을,
feature-platform은 실행부가 같이 쓰는 부팅 모듈을 한 청크로 묶는다.
render-boot-failure.tsx가 쓰는 모듈은 feature-platform에서 뺀다(09-03:
애플리케이션 그래프가 평가 중 던져도 실패 화면은 떠야 한다).
버린 대안
- 부팅 때 feature 청크를 미리 동적 import하고 기다린 뒤 렌더: 분류기는
lazy로 세지만 첫 렌더 경로의 바이트는 그대로다. 예산을 속이는 것이라
버렸다.
- 실행부 일부만 지연(adapter와 contribution만): 모두 같은 zod 스키마를
공유해 줄지 않았다. 실측 +4,031 B로 오히려 늘었다.
- 선언과 실행부를 같은 lazy 청크(context-routes)에 두기: 기능당
-60 B 수준이라 그룹 없이 목표에 못 미쳤다.
- 페이지 문구까지 실행부로 옮기기: 기능당 약 300~450 B를 더 줄이지만
i18n catalog를 런타임 등록 구조로 바꿔야 해 이번 범위에서 뺐다.
- boot.tsx를 별도 진입점으로 추가: 분류기가 boot의 동적 import를 모두
초기로 세어 218,303 B가 됐다.
위험·한계
- FeatureModule 모양이 바뀐다. 다운스트림 기능은 feature-runtime.ts로
실행부를 옮겨야 한다(BREAKING CHANGE).
- 기능 계약 간 중복(operation, schema, contribution) 검사가 부팅이 아니라
그 기능의 적재 시점과 Node 검사(loaded-feature-runtimes를 읽는 스크립트·
테스트)로 옮겨 간다.
- 기능 route에 처음 들어갈 때 실행부 청크(약 5 KB gzip)를 받는다. 플랫폼
구성 화면은 모든 기능의 실행부를 받은 뒤 그린다. lab 게이트(FE-GATE-026)가
재는 화면이라 LCP가 조금 늘 수 있다. 측정하지 않았다.
- 청크 그룹은 rolldown 자동 분할을 일부 대신한다. feature-declarations가
비면 기능당 증가 상한 테스트가 실패한다. feature-platform이 비면 템플릿
총량(약 2 KB)만 늘고 번들 예산 검사가 잡는다.
- config/contracts/registry-baseline.json의 FE-REG-API/SCHEMA source는 아직
installed-feature-contracts.ts다. 승인 digest가 걸려 있어 소유자 재승인
때 고친다. check-registries는 통과한다.
- 예산 안에 들어가는 참조 모양 기능은 약 3개다. 플랫폼 자체가 예산의
98.8%다.
수치 (vite build, classifyViteJavascript + gzipSync 기본 수준)
- 템플릿: 203,572 B(13파일) -> 202,361 B(16파일), -1,211 B
- 참조 복사본 +1: 207,494 B(+3,922) -> 203,102 B(+741)
- 참조 복사본 +2: 210,818 B(+3,324) -> 203,750 B(+648)
- 참조 복사본 +5: 205,629 B
- lazy JS: 23,516 B -> 28,204 B
검증
- RED: node scripts/run-vitest.ts run tests/unit/feature-shape-growth.test.ts
"adds at most 1 KiB ... (15-01)" 실패: expected 3934 to be less than or
equal to 1024
- GREEN: 같은 테스트 통과(+741 B)
- node scripts/run-vitest.ts run tests/unit: 168파일 2,383 통과
- node scripts/run-vitest.ts run tests/component tests/features
tests/integration tests/contract tests/runtime-schema tests/architecture
tests/platform tests/recipes (security-followup-archive 제외, pnpm 호출):
109파일 787 통과
- npx tsc --noEmit: app, node, test, recipes, web-worker, service-worker 통과
- npx eslint 바뀐 파일 통과
- check-architecture.ts(pnpm exec 대신 eslint를 부르는 shim): PASS, 새 규칙이
boot.tsx의 import를 거절하는 것 확인
- check-adapter-inventory, check-registries, verify-documentation-readiness,
check-ci-contract: PASS
- e2e Chromium 전체(pnpm 없는 build-frontend + vite preview, 별도 포트):
55 통과
- storybook build 성공
- 미실행: Firefox·WebKit e2e, 시각 회귀, lab 성능(FE-GATE-026),
pnpm 경로의 check:bundle, test:system
BREAKING CHANGE: FeatureModule은 id, contracts(선언), routes, messages,
loadRuntime만 가진다. adapters, contractContributions와 V2 API 계약
(apiOperations, schemas, runtimeSchemas, mappers)은 defineFeatureRuntime으로
만든 feature-runtime.ts로, route codec은 context-routes.ts의 route handle
codecs로 옮긴다. contracts에는 contractPackages를 추가한다.
Refs: docs/reviews/2026-11-01-maturity-review/15-performance.md 15-01,
16-structure-patterns.md 9절
FE-GATE-012(build, check-bundle, check-app-build-input-bundle,
check-optional-recipes)는 release 단계에만 있어 tag push나 수동 release
실행에서만 돈다. 초기 JS를 예산보다 크게 만드는 PR도 merge 게이트를
통과하고, 몇 PR 뒤 release가 실패한다. 어느 PR이 얼마나 늘렸는지는 그때
이력을 거꾸로 따라가 찾아야 한다.
FE-GATE-012를 merge stage와 merge_gate job으로 옮긴다. release_gate는
merge_gate를 needs로 두므로 tag 실행에서도 그대로 돈다. promotion 공식
(MERGE_READY)과 CI 계약 검사의 job 소유 표를 같이 바꾸고, workflow는
생성기(node scripts/generate-ci-workflow.ts)로 다시 만든다.
버린 대안:
- check-bundle만 FE-GATE-011(build)에 붙인다(리뷰의 최소안). 같은 검사가
두 게이트에 갈라지고, release에 남은 FE-GATE-012와 중복해서 돈다.
- FE-GATE-012를 merge와 release 둘 다에 둔다. 계약이 게이트당 job 하나를
강제한다(multiply owned gate).
- lab 게이트(FE-GATE-026)도 merge로 옮긴다. PR마다 브라우저 설치와 5회
측정이 붙고, 부하에 따라 판정이 뒤집힌 기록(0929/11-04)이 있다. release에
남긴다.
위험·한계:
- merge 실행이 build 1회와 공급망 생성, 번들 계산만큼 길어진다.
- 실제 CI 러너에서는 돌려 보지 않았다(Actions 꺼짐).
검증:
- node scripts/check-ci-contract.ts: 27 gates PASS
- node scripts/generate-ci-workflow.ts --check: PASS
- node scripts/generate-artifact-schemas.ts --check: PASS
- node scripts/run-vitest.ts run tests/unit/ci-workflow-generation.test.ts
tests/unit/promotion-readiness.test.ts: 423 통과(workflow 스냅샷 1건 갱신)
Refs: docs/reviews/2026-11-01-maturity-review/15-performance.md 15-01
FE-GATE-021..025 rebuilt from source and drilled that build, so the drilled
bytes were not the promoted release. production_gate now downloads the
promoted-release artifact, extracts it against immutable_build's digest and
runs each drill with --release-root; the CI contract rejects a rebuild.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
check-ci-contract only caught test files outside the gates; check:* scripts
such as check:request-fanout and check:remediation-ledger ran nowhere. The
new check reads config/ci/ungated-scripts.json; three checks are now wired
into FE-GATE-010 and FE-GATE-012 and the rest carry a reason.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The growth test read docs/reviews/, which the removal fixtures do not copy and
README lets a project delete, so three removal checks failed with ENOENT. The
rationale now lives in config/scale/feature-shape-ceilings.json and a guard
test keeps code from reading the deletable history documents.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The budget is a size ratchet set about 100 bytes above the measured
reference bundle. file-transfer 55582 -> 55623 gzip bytes: 01-07 moves
the vault release into one finally around put() and open(). Same
rounding rule: 55700.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The budget is a size ratchet set about 100 bytes above the measured
reference bundle. Measured after merging lanes B, D, F, G, H and J:
realtime 39903 -> 40886 gzip bytes (04-01..04-08 add reconnect rules and
SSE handling; 06-01, 06-02, 06-06, 06-09 add web-push fences and click
handling, and src/adapters/web-push is a realtime recipe source root).
Same rounding rule: 40900.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The detail path /examples/reference-resources/:resourceId lost to the static
new and status routes, so an item with one of those ids opened the wrong
screen. The breaking route change carries registry evidence.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The scale evidence came from stub features, so the cost of one app growing
towards its declared ceiling was never measured. Copies of the reference
feature at 25 and 100 are now the baseline, and the acceptance matrix says
what the ceilings rest on and which axes stay unverified.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ENV_REGISTRY called VITE_BUILD_ID required with no default while every reader
fell back to local-build. Staging and production builds now fail without it;
local builds keep the placeholder, and the registry records that default.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The policy guarded legacy retry-policy.ts but not the live V3 retry and
contract path rules, and the strategy doc claimed 19 modules at 90% targets.
The doc now states the enforced ratchet. The below-threshold fixture used the
pre-counter summary shape and failed with a TypeError instead of the
'Risk coverage failed:' diagnostic the negative gate expects.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The gate only matched one diagnostic string, so widening the telemetry event
name or the diagnostic level still counted as the expected failure. Each
violation now has its own directive and the fixture must compile cleanly.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nothing read severity, telemetryEvent or redaction, yet FE-REG-ERROR guarded
them as policy. They are removed and the registry baseline is re-approved.
A new test fails if createFailure stops clearing retryable for MAYBE_APPLIED (M1).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The classifier called type, enum and nullable changes "none" and treated new
keys under strict readers as additive; FE-GATE-014 only checked hand-written
fixtures. It now fails closed and compares the committed baseline with the
schema the build emits. The doc lists per-boundary reader rules.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The runtime does not know the session, so File refs, preview leases and
downloads of the previous user outlived a logout. Section 3.5 and the recipe
lifecycle now say to dispose it, via a scope reset participant if needed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The budgets are size ratchets set about 100 bytes above the measured
reference bundle. Measured on 926f448 and after the wave-1 lanes:
file-transfer 54468 -> 55517 gzip bytes (01-01..01-11 add capability
reissue, size-aware timeouts, probe limiting) and offline-indexeddb
35856 -> 36822 (05-01..05-14 add per-transaction locks, purge sweep and
corrupt-row skipping). Same rounding rule: 55600 and 36900.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A feature directory without feature-module.ts vanished silently, registry
consumers listed only the sample's directories, and the removal gate only knew
the sample. The generator now fails unless a NOT_INSTALLED marker says so
(D-10), consumer directories are globs over every feature, and the removal
script takes --feature=<id>. starter-experience.md documents the current
ownership declarations and the files a new feature touches.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
FE-GATE-019 ran before the release existed and passed by comparing the declared
headers with a copy of themselves. The fixture result is now UNVERIFIED, the
gate runs in the production stage after deployment, and the live probe
requests the entry script the served index.html names (D-14).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
config/architecture/layers.json was read by nothing and contradicted the
enforced adapter rule. It is deleted; .dependency-cruiser.json is the one rule
source, and the ESLint layer patterns are marked as an editor aid.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
No CI job ran any scale tier, and every tier past pr compares timings only on
the baseline CPU. The pr tier checks counts, bounds and growth shape; it now
includes S2 composition, since S0/S1 stay under the noise floor and let a
quadratic accumulation through. The ungated-test check reads tier paths.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Drills held constant true verdicts and read public/ placeholders, so they
passed with no build or with JS assets deleted. They now check the local dist,
mark boot, live traffic and provider steps UNVERIFIED (exit 3), and FE-GATE-016
is retired because it repeated FE-GATE-025's FE-RB-005 drill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The signed review is per release candidate, so the merge gate could never pass.
FE-GATE-009 now checks only the record format; the checked-in records are
source-controlled evidence, not command output.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tests/architecture, tests/release, tests/scripts, the top-level
tests/features files and the browser-soak contract now run in
test:coverage (FE-GATE-005); check:adapter-inventory joins FE-GATE-010
and the image resolve-signal type fixture joins FE-GATE-003 (baseline
updated). check:ci fails on a test file no gate command reaches unless
config/ci/ungated-tests.json gives a reason; only the tests/scale suite
is listed, pending 11-03. defineFeatureModule's id mismatch is tested.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Gate/command/artifact counts and the gate-shape SHA-256 were code
constants, so any real removal or addition failed check:ci and removal
fixtures passed only in a relaxed mode. They now live in
config/ci/contract-baseline.json (owner, reason) with
update:ci-contract-baseline; update:registry-baseline wraps the existing
registry script. Removal fixtures re-baseline and run canonical check:ci
and check:registries; CI_CONTRACT_MODE, --reduced-removal-fixture and
check:registries:structure are gone. ci-quality-gates.md lists the
baselines a new project resets.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
verify-hosting-headers.ts passed the serviceWorker probe only when
cache-policy.json said `enabled: false` AND dist had no worker script, so no
configuration let a supported ACTIVE build through the hosting gate. The
probe now follows the static selection: ACTIVE requires exactly the selected
`service-worker.js` at the dist root; null / REMOVE_REGISTRATION /
PURGE_OWNED_RESOURCES still reject any worker script. The now-unused
`serviceWorker.enabled` policy flag is removed; the artifact keeps the same
`serviceWorker`/`enabled` probe key.
RED (ACTIVE build from the previous commit, `node scripts/verify-hosting-headers.ts`):
Hosting cache/content-type/security header verification failed. exit=1
{ surface: serviceWorker, header: enabled, expected: false, observed: true, passed: false }
GREEN: same ACTIVE build -> PASS exit=0 (expected: true, observed: true);
default build -> PASS; default build + stray dist/service-worker.js -> exit=1
(reason "dist worker scripts: service-worker.js");
tests/unit/hosting-probe.test.ts 13 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Delete src/features/reference-feature/presentation/reference-feature-runtime.tsx:
it only re-exported the same-named .ts and nothing imported it by the
.tsx path.
- Move promotion-readiness.ts and performance-budgets.ts from
src/application/policies to scripts/lib: only scripts and their unit
tests import them. Imports updated in 5 scripts and 2 tests.
config/testing/risk-coverage.json lists only src modules, so their two
criticalModules/highRiskPaths entries are removed; the synthetic
14-file summary fixture now names two other existing src modules.
- installed-feature-messages.ts composes one catalog per locale the
installed features declare (composeInstalledMessageCatalogs) instead
of a fixed ko-KR/en-US list that silently dropped other catalogs. The
exported type still maps each declared locale to its exact key set.
Verified: tests/unit/installed-feature-messages.test.ts (ko-KR/en-US
content pinned first; derived-locale and duplicate-key cases RED before);
app typecheck also passes with a zero-feature generated index; all gates
PASS; check:ci PASS; check-bundle resolves the moved module (stops later
on the absent artifacts/performance/bundle.json).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>