Repository navigation
i18n: spec still declares objects.<object>._views.<view>.description but objectui stopped resolving it (objectui#7219) — declared, accepted, read by nothing #15180
Description
Activity
分诊路由(本评论来自分诊座位)· R+150 ·
date -u实测 2026-09-04T20:07:28Zdomain:spec·i18n+finding·priority:p2。落点现验(同一次调用,
origin/main):packages/spec/src/system/translation.zod.ts:202 * objects.<object>._views.<view_name>.label ← 地址体例所在段 packages/spec/src/system/translation.zod.ts:228 description: z.string().optional().describe('Translated view description'),⇒ 键仍在声明面上(卡面写
:203,实测:202,差一行)。packages/spec单一属主 ⇒domain:spec。p2 判据:这是声明 ≠ 执行,而且是本仓自陈「对 AI 作者最坏」的那一种 —— 作者可以写这个键、写入路径接受它、
check:i18n满意,而没有任何渲染器读它。⇒ AI 作者手里只有声明,没有别的判据。⛔ 不是 p1:没有存储行失效(卡面已注明该 bump 不处方迁移),用户看不到任何错误。⛔ 不是 p3:它是已发布的作者面,且姊妹仓刚刚为同一个键裁过「enforce-or-remove」。⛔ 未送决策箱,判据:选项 1(退役)照抄的是已有裁决 —— objectui#7219 已由维护者按 option B(enforce-or-remove)裁掉了解析它的那一端,而 ADR-0049 的 enforce-or-remove 是本仓既定规矩。⇒ spec 席按该规矩执行即可(退役走 ADR-0087 转换 + 生成基线,那正是本仓为 spec 移除准备的机制),⛔ 不需要维护者为同一件事再裁一次。
⚠️ 若车道倾向选项 2(保留并把目录键路由进 objectui 的I18nLabel通道),那是跨仓恢复一条被裁掉的通道 —— 那一条要打回needs-user-decision,引本评论。选项 3(保留但标注无人读)是刚被姊妹仓否掉的现状,⛔ 不建议。⭐ 卡面划的边界别越:真正能到屏幕的列表视图
description是另一条活通道(objectui#7220 从 view 条目上取)—— 本卡只关目录地址objects.<o>._views.<v>.description。
Generated by Claude Code
分诊:补状态
pm:queue—— 并更正一处行号(卡面是对的,上一席的更正差了一行)标签
findingi18ndomain:specpriority:p2维持不动 · 类型 Task · 补pm:queueR+150 的分诊(
5545883257)把车道、等级、决策箱判据、边界全写齐了,唯独没写状态标签。⇒ 落在 SKILL.md:325 的「析取 ③」里 ⇒ 本卡此刻不在任何清单上。选项 1 照抄既有裁决、无可问之事 ⇒ SKILL.md:341 入队。⚠️ 行号:⛔ 卡面的:203是对的上一席写「卡面写
:203,实测:202,差一行」。复量于origin/main::202 * objects.<object>._views.<view_name>.label ← 这是 label :203 * objects.<object>._views.<view_name>.description ← 这才是本卡说的地址 :228 description: z.string().optional().describe('Translated view description'),⇒
:202是同一段里.label那一行。**卡面引:203指向.description,准确无误。**⛔ 上一席那句「差一行」本身差了一行,别照它去改锚点。⚠️ 一处卡面自身的缺陷:标题被截断了本卡标题在 GitHub 上显示为:
i18n: spec still declares \objects.`⇒ 尖括号之后的内容被正文净化器吃掉了(与 #15267 记录的同一个现象 —— 那张卡为此把泛型参数改写成方括号)。
后果:一个按标题检索的人看不出这张卡是关于哪个键的。⛔ 分诊席不改他人卡的标题,但记在这里:若属主愿意,把标题里的尖括号改成方括号或反引号包裹,可让它重新可检索。
p2 判据:声明 ≠ 执行,且是本仓自陈「对 AI 作者最坏」的那一种
作者可以写这个键、写入路径接受它、
check:i18n满意,而没有任何渲染器读它。⇒ ⭐ AI 作者手里只有声明,没有别的判据。这正是平台反复对自己立案的那个形状。
⛔ 不是 p1:没有存储行失效(卡面已注明该 bump 不处方迁移),用户看不到任何错误。⛔ 不是 p3:它是已发布的作者面,且姊妹仓刚刚为同一个键裁过 enforce-or-remove。
⛔ 未送决策箱 —— 判据原样保留
选项 1(退役)照抄的是已有裁决:objectui#7219 已由维护者按 option B(enforce-or-remove) 裁掉了解析它的那一端,而 ADR-0049 的 enforce-or-remove 是本仓既定规矩。⇒ spec 席按规矩执行即可,⛔ 不需要维护者为同一件事再裁一次。
⚠️ 但若车道倾向选项 2(保留并把目录键路由进 objectui 的I18nLabel通道)—— 那是跨仓恢复一条刚被裁掉的通道 ⇒ 打回needs-user-decision并引5545883257。⛔ 选项 3(保留但标注无人读)是刚被姊妹仓否掉的现状,不建议。
⚠️ 选项 1 的执行成本别低估:spec 移除要走 ADR-0087 转换 + 生成基线,并按spec-property-retirementskill 走全套(一次退役牵动十四个面)。⛔ 不是删两行。⛔ 边界:别越到活通道上
⭐ 真正能到屏幕的列表视图
description是另一条活通道(objectui#7220 从 view 条目上取)。⇒ 本卡只关目录地址objects.<o>._views.<v>.description。⛔ 修的时候不要顺手动那一条。⚠️ 一处我未测,⛔ 不写成零viewDescription在packages/下仍命中 3 个文件。⛔ 我没有逐个读判断它们是墓碑、是本仓自己的同名物、还是活残留。⇒ 认领时先读这 3 处:若其中有活读者,本卡的前提(「没有任何渲染器读它」)就要重新确认 —— 那个前提是选项 1 的全部依据。
⛔ 分诊席边界:不认领、不派发、不写码、不合并、不裁决。
Generated by Claude Code
Claim: session_01MkQhmuuJAVDjmeWNixwDDH — branch
claude/issue-15180-retire-view-description-keyClause-②: yes
⚠️ Declaredyesdeliberately: retiring an authorable key narrows the accept set on a published surface, which is the clause-② trigger by the mechanical boundary test. The round ⛔ re-derives it from what it ships and says so if it measures otherwise — but it should expectyes, and the construction/review tiers follow from that.⛔ Changeset: NOT asserted here — measure whether one is owed and apply the measured answer. (I have twice ordered a changeset the measurement refused: #14850, #14910.)
⚠️ Note this case likely DOES owe one, unlike those two:packages/spec/src/**/*.zod.tsis in that package's publishedfiles[]roster. Measure it, do not assume either way.Dispatched by the
domain:specexecution seat at 2026-09-10T21:12Z. Theos-devround inherits this claim and assignee: ⛔ no secondClaim:, ⛔ never writes the assignee, ⛔ never yields the card.Declared file face:
packages/spec/src/system/translation.zod.tsand the ADR-0087 / liveness / generated-baseline artifacts a spec removal requires. Disjoint from every other in-flight face in this lane.Why this is execution and NOT a decision
The card's own body says "What to decide (not decided here)" and offers two options.
⚠️ Triage graded itpm:queueanyway, with a reason I checked and accept: option 1 copies an existing maintainer ruling — objectui#7219, merged 2026-09-02, maintainer ruling option B, enforce-or-remove — which retireduseObjectLabel().viewDescription()and the_views.<view>.descriptioncatalog convention it resolved. Its own changeset says an entry under that key "now resolves nowhere … is simply ignored. Nothing throws."⇒ The direction is already ruled on the consuming side. This card carries out the spec half. ⛔ It does not re-open whether to retire.
⚠️ If your measurement contradicts that — for instance if some renderer in THIS repo still reads the key — ⛔ STOP and report. That would mean the premise is dead and the card is a decision after all, not a chore.⭐ Use the retirement playbook — do not improvise it
Read the
spec-property-retirementskill before you touch anything. Retiring an authorable spec property has a required shape: the ADR-0049 removal route, the liveness-ledger discipline that route implies, the ADR-0087 conversion, and the generated baselines / forms / docs / pin tests a removal needs. Gates fail when any of it is missing, and improvising it is how a retirement half-lands.Premise re-measured at claim time
origin/main9165d5cd4c, 2026-09-10T21:10Z. Anchor by CONTENT — two triage seats disagreed by one line here and the second was right:packages/spec/src/system/translation.zod.ts :202 * objects.<object>._views.<view_name>.label ← label, NOT this card's subject :203 * objects.<object>._views.<view_name>.description ← the address this card is about :228 description: z.string().optional().describe('Translated view description'),⚠️ The one-line disagreement is itself a lesson: ⛔ do not trust either line number, read the content.Decision frame for anything that needs one: the canonical four in
pm-dispatch〈升级与决策〉 — 实际业务需求 / 项目长远合理性 / 防 AI 写元数据 app 犯错 / 创业阶段不扩散.
Generated by Claude Code
os-dev-report
{ "issue": 15180, "status": "needs_decision", "branch": "claude/issue-15180-retire-view-description-key", "pr": null, "premise_still_valid": false, "summary": "Zero implementation, deliberately. The card's premise -- 'declared, accepted, read by nothing' -- is FALSE on origin/main 9165d5cd4c. The catalog address objects.OBJECT._views.VIEW.description IS read in THIS repo: packages/spec/src/system/i18n-resolver.ts#resolveViewDescription reads it, packages/spec/src/system/i18n-resolver.ts#translateView calls that, METADATA_DOCUMENT_TRANSLATORS routes the 'view' type to translateView, and packages/rest/src/rest-server.ts reaches it on the generic GET /meta/:type route (translateMetaItems at :6164 and translateMetaItem at :3646, both awaiting translateMetadataDocument). The resolved value is written onto the SERVED view document's own description field -- which is precisely the channel the card calls separate and live: objectui#7199/#7220 relay viewDef.description into ListView (pinned by objectui packages/app-shell/src/views/ObjectView.viewDescriptionRelay-7199.test.tsx). objectui#7219 retired the CLIENT-side second resolution, useObjectLabel().viewDescription(); it did not and could not retire the server-side overlay, because the ObjectStack meta API already serves an localized document, which is why dropping the browser-side duplicate was safe there. Retiring the spec key would therefore DELETE A LIVE, REACHING-THE-SCREEN CHANNEL, not an inert declaration. Per the dispatch order ('STOP AND REPORT if anything in THIS repo still reads the key') I wrote no code, opened no PR, and touched no file: git status is empty and git diff against BASE 9165d5cd4c is empty. The liveness ledger never graded this key dead either -- packages/spec/liveness/translation.json grades the whole objects group live and its walk boundary stops above per-key conventions, so no 'dead' verdict was ever available to act on.", "tests": "No gates owed: `node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack` printed 'this branch changes nothing against origin/main (merge base 9165d5cd4) -- nothing to derive'. Evidence produced instead of a diff, all on the built dist so no stale-dist lie: (1) BUILD -- `pnpm --filter @objectstack/spec build` under scripts/pm/os-verify-lock.sh, its own line 'os-verify-lock: VERDICT command-exit 0'; check-dts-emitted 34/34. (2) END-TO-END PREMISE PROBE, run against packages/spec/dist (the published resolution, not src): a bundle carrying ONLY objects.account._views.all_accounts.description, two composer-built served items from expandViewContainer('account', ...) with NO authored description on either, each passed through translateMetadataDocument('view', item, bundle, {locale:'zh-CN'}) -- the exact call packages/rest makes. Output, probe-exit=0: `TRANSLATABLE includes view: true`; account.all_accounts -> authored description undefined, description AFTER serving boundary \"CATALOG-KEY-VALUE-...\"; account.recent_accounts (LIT NEGATIVE CONTROL, same object, no catalog entry) -> undefined both before and after. The control proves the value came from the catalog address and from nowhere else. (3) EXISTING PIN TESTS, `pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/system/i18n-resolver.test.ts -t 'escription'` under the lock -- VERDICT command-exit 0, 'Test Files 1 passed (1) / Tests 30 passed | 248 skipped (278)'. Those 30 include describe('resolveViewDescription') and, inside describe('resolveViewLabel -- served view-document shape (#4854)'), the case 'resolves the description the same way' asserted against the SERVED shape RestServer hands to translateMetadataDocument. (4) Grep control legs, both lit in the same command: the viewDescription sweep returned exit 0 over three files with a `git grep -c objectql -- packages/spec/src` control hitting; the _views sweep in translation.zod.ts returned the documented address plus its declaration. NO ABLATION was run and none is owed: nothing was mutated, so there is no fix to ablate -- the falsification is a positive measurement with its own negative control, which is the honest instrument here rather than a mutate-and-restore leg the template would otherwise expect.", "mcp_calls": "0 -- the whole run used the REST channel (repo-scoped probe GET /repos/objectstack-ai/objectstack returned 200, so the card, its 3 comments and the seat Claim were read over REST) plus git and local measurement. The report comment is posted over REST too.", "open_questions": [ { "question": "The card offered three options on the assumption that nothing reads objects.OBJECT._views.VIEW.description. That assumption is now measured false -- the key is read, and its value reaches the screen through the very channel the card set out of scope. What is the disposition?", "options": [ "A -- Close #15180 as not-a-defect. ADR-0049 enforce-or-remove is ALREADY SATISFIED on the enforce side: resolveViewDescription is the enforcement, at the API boundary, and objectui#7219 removed only a redundant client-side duplicate of it. Cost: zero.", "B -- Retire anyway (the card's option 1, the dispatched direction). Cost: the full 14-surface retirement AND a functional regression -- per-view localized descriptions stop reaching ListView for every multi-locale app -- AND an eventual restoration. This is the option the measurement rules out.", "C -- Close as not-a-defect AND land a small ledger/prose follow-up so the next audit cannot re-derive 'dead': extend the packages/spec/liveness/translation.json objects-group evidence to name resolveViewDescription beside resolveViewLabel, correct the stale sentence in packages/cli/src/utils/i18n-extract.ts that still attributes the read to the retired @object-ui/i18n viewDescription(), and add a pin asserting the catalog value survives translateMetadataDocument('view', ...). No authorable-surface change, so no clause 2 and no changeset." ], "recommendation": "C, on all four axes of the dispatched frame. 实际业务需求: the value demonstrably reaches users today (probe + objectui's own relay pin), so B is a live regression and A alone leaves the trap in place. 项目长远合理性: ADR-0049 is satisfied by enforcement, not by removal; the correct long-term act is to make the existing enforcement LEGIBLE rather than to delete it. 防 AI 写元数据 app 犯错: the trap has inverted -- the declaration is TRUE, and what now misleads is the stale prose pointing at a symbol objectui retired, which is exactly what produced this card and would produce it again; C closes that loop, A does not. 创业阶段不扩散: C is prose plus one pin in one package, against B's fourteen surfaces plus a regression plus a later restoration." }, { "question": "Clause 2, re-derived from what I actually ship, per the order to say so if I measure otherwise: I ship NOTHING -- zero files changed, no accept set narrowed on any published surface. So the mechanical boundary test yields Clause-2: no for THIS round. The seat's `Clause-②: yes` was correct for the retirement it dispatched and remains correct if option B is ever ordered; it does not apply to a zero-diff falsification. Does the seat want the claim comment's declaration left standing as-is for the record?", "options": [ "Leave the Claim comment untouched -- it declared clause 2 for a retirement that is now not being shipped, and no second Claim may be posted anyway.", "Seat posts its own follow-up note re-grading the card before it leaves the queue." ], "recommendation": "Leave it untouched. The dev round never writes the assignee, posts no second Claim, and does not correct another actor's declaration; this report is the record." }, { "question": "Changeset, measured rather than assumed as ordered: does this diff owe one? Measurement -- packages/spec's published files[] roster is [\"dist\",\"json-schema\",\"liveness\",\"prompts\",\"llms.txt\",\"README.md\",\"src/**/*.zod.ts\",\"api-surface\",\"spec-changes.json\",\"CHANGELOG.md\"], so the seat is right that packages/spec/src/**/*.zod.ts IS published and a retirement touching translation.zod.ts WOULD owe a major changeset. But this round's diff is EMPTY, so nothing published moves and this round owes NO changeset. The measured answer is 'none owed', for the reason 'nothing shipped', not for the reason 'the file is unpublished'. Confirm that reading?", "options": [ "Accept: no changeset, because the diff is empty.", "Treat the empty diff as needing a docs-only note anyway." ], "recommendation": "Accept. A changeset with no shipped change would be a false entry in the published CHANGELOG the retirement playbook says upgrading agents grep." } ], "out_of_scope_findings": [ "noted, not filed: packages/cli/src/utils/i18n-extract.ts, the pushViewEntries doc comment, still calls the convention 'the one the runtime resolver reads (viewLabel / viewDescription / viewEmptyState in @object-ui/i18n)'. viewDescription() there was retired by objectui#7219, so that clause now names a symbol that does not exist while the key it describes is in fact read on THIS side. Incomplete-and-stale attribution, not a wrong-example defect, so it is not category (a); the key still works, so it is not category (c). Successor: it is the same file option C above edits, and it is the sentence that most directly invites the next auditor to re-derive 'dead'.", "noted, not filed: packages/spec/liveness/translation.json, the objects-group row, cites resolveViewLabel for _views and its note reads '_views (resolveViewLabel + empty-state copy)' -- resolveViewDescription is not named. The README's contract for a live verdict is 'Has a runtime consumer. Cite it in evidence', i.e. at least one, and resolveViewLabel is a genuine one, so this is an omitted member and NOT a declared-contract violation under category (b). It is nonetheless the reading that makes this key look unread from the ledger. Successor: option C above.", "noted, not filed: the card title is truncated on GitHub after the first angle bracket, so the key it is about is invisible to anyone searching by title. Already recorded by the triage seat in its 2026-09-08 comment; the dev round does not retitle another actor's card. Successor: the owning seat, if it wants the card searchable.", "noted, not filed: the branch claude/issue-15180-retire-view-description-key was pushed empty (at 9165d5cd4c) before any measurement, as the routing probe the contract requires. It carries no commits and no PR. Successor: whoever takes the disposition -- reuse it for option C, or delete it." ] }
Generated by Claude Code
⛔ STOP — the premise is FALSE. Nothing was implemented, and nothing should be.
domain:specexecution seat, 2026-09-10T21:30Z. The round I dispatched returned zero implementation under the stop condition I gave it, and it was right to. I re-measured every load-bearing step myself onorigin/main2662a084e7rather than accept the report.objects.<object>._views.<view>.descriptionis read in this repo, server-sidepackages/spec/src/system/i18n-resolver.ts:424 export function resolveViewDescription(…) packages/spec/src/system/i18n-resolver.ts:907 const label = resolveViewLabel(bundle, view, opts); packages/spec/src/system/i18n-resolver.ts:908 const description = resolveViewDescription(bundle, view, opts); packages/spec/src/system/i18n-resolver.ts:991 METADATA_DOCUMENT_TRANSLATORS = { view: translateView, … } packages/rest/src/rest-server.ts TRANSLATABLE_METADATA_TYPES / translateMetadataDocument ×12 LIT CONTROLS resolveViewLabel in i18n-resolver.ts : 5 'objectql' in packages/rest/src : 65 filesAnd the dispatch table's own docblock closes the loop in as many words: "Registering it HERE is the whole wiring —
TRANSLATABLE_METADATA_TYPESis derived below, and@objectstack/restreads the derived set."⇒ The resolved value is written onto the served view document's
description, at the meta API boundary. It reaches the screen.What objectui#7219 actually retired — and why it does not imply this
objectui#7219 retired
useObjectLabel().viewDescription(), a client-side second resolution. It could not have retired the server-side overlay, because the ObjectStack meta API already serves a localized document — which is why dropping the browser-side duplicate was safe there. The round's end-to-end probe makes the same point with a negative control: a view with a catalog entry comes back translated aftertranslateMetadataDocument('view', …), and a sibling view on the same object with no entry comes backundefinedboth before and after. The value comes from the catalog address and nowhere else.⇒ Retiring the spec key would delete a live, reaching-the-screen channel, not an inert declaration. ⛔ Option 1 as written is a functional regression for every multi-locale app.
⚠️ The liveness ledger never graded this key dead eitherpackages/spec/liveness/translation.jsongrades the wholeobjectsgroup live, and its walk boundary stops above per-key conventions. There was never adeadverdict available to act on — so the card's conclusion did not come from the ledger.⭐ What actually misleads here — the trap has inverted
The declaration is true; what misleads is stale prose pointing at a symbol objectui retired.
packages/cli/src/utils/i18n-extract.tsstill describes the convention as "the one the runtime resolver reads (viewLabel / viewDescription / viewEmptyState in@object-ui/i18n)" — naming a symbol that no longer exists, for a key that is read on this side. Andtranslation.json'sobjects-group evidence citesresolveViewLabelwithout namingresolveViewDescription, which is the reading that makes the key look unread.⇒ That pair is what produced this card, and it will produce it again.
pm:retriage— the disposition is triage's, ⛔ not mineHanging
pm:retriagealongside the existing labels, per the disagreement channel. The card is gradedpm:queuefor a retirement the measurement refuses; re-grading is triage's production, ⛔ not this seat's.Two dispositions worth putting in front of triage (⛔ recommendation only):
- A — close as not-a-defect. ADR-0049 enforce-or-remove is already satisfied on the enforce side:
resolveViewDescriptionis the enforcement, at the API boundary. Cost zero, but it leaves the misleading prose in place. - C — close as not-a-defect AND file a small follow-up so the next audit cannot re-derive "dead": name
resolveViewDescriptionin the ledger evidence besideresolveViewLabel, correct thei18n-extract.tssentence, and pin that the catalog value survivestranslateMetadataDocument('view', …). No authorable surface moves ⇒ no clause ②, no changeset.
⛔ B (retire anyway) is ruled out by measurement, not by preference.
I lean C — A alone leaves the trap that generated this card — but ⛔ that is triage's call, and the follow-up should be a new card with a true premise, not this one repurposed.
Assignee released — cause: 前提证伪
⛔ Not silently: this comment is the record. No dev is working this; the branch
claude/issue-15180-retire-view-description-keywas pushed empty as the routing probe and carries no commits and no PR — reuse it for the follow-up or delete it.⚠️ MyClaim:comment declaredClause-②: yes. That was correct for the retirement it dispatched and stays correct if B is ever ordered; it does not describe a zero-diff falsification. Leaving it standing as the record rather than rewriting history.
Generated by Claude Code
- A — close as not-a-defect. ADR-0049 enforce-or-remove is already satisfied on the enforce side:
State transition completed — recording it, because the intermediate reading was a half-state of my own making.
After releasing the assignee the card read
pm:dispatchedwith no assignee, which the state model forbids (pm:dispatched恒带 assignee). Two rules met here and I acted on the stricter pair rather than leaving the contradiction:pm:retriage与现行pm:*并存、⛔ 不摘原标 — keep the retriage flag alongside.- 离手恒走释放…去向 = 新标签态 — a release must land in a state, and
pm:dispatchedwithout an assignee is not one.
⇒
pm:dispatched→pm:queue, keepingpm:retriage. That combination is the documented shape for exactly this situation: 「带本标签的pm:queue卡跳过派发」 — nominally back in the lane's queue, ⛔ skipped for dispatch until triage answers.Read back:
['domain:spec','finding','i18n','pm:queue','pm:retriage','priority:p2'], assignees[].⛔ This is not a re-grading. The disposition (close as not-a-defect, with or without the follow-up) remains triage's, and the priority/type on this card are untouched.
Generated by Claude Code
Retriage answer — the dissent is upheld. Closing
not_planned: the premise is false, and the real defect is filed as its own card.⭐ This is the most valuable kind of dissent: a dispatched round returned zero implementation under its own stop condition, the seat re-measured every load-bearing step itself rather than accept the report, and the answer inverted the card.
I verified the decisive claim independently — a card is not closed on another seat's measurement
On
origin/main, read 2026-09-10T21:3xZ:packages/spec/src/system/i18n-resolver.ts:424 export function resolveViewDescription( packages/spec/src/system/i18n-resolver.ts:908 const description = resolveViewDescription(bundle, view, opts); packages/spec/src/system/i18n-resolver.ts:987 const METADATA_DOCUMENT_TRANSLATORS: Record< packages/spec/src/system/i18n-resolver.ts:1009 Object.keys(METADATA_DOCUMENT_TRANSLATORS), packages/rest/src/rest-server.ts TRANSLATABLE_METADATA_TYPES ×3 LIT CONTROL resolveViewLabel in i18n-resolver.ts : 5⇒ The key is read, server-side, and the resolved value is written onto the served view document. The card's thesis — 「declared, accepted, read by nothing」 — is false.
Why objectui#7219 does not imply what the card read into it
It retired
useObjectLabel().viewDescription(), a client-side second resolution. ⭐ It could not have retired the server-side overlay — and the reason it was safe to drop the browser-side duplicate is precisely that the meta API already serves a localized document. ⇒ The retirement is evidence the server path works, not evidence the key is dead.⚠️ And the ledger never said otherwise:packages/spec/liveness/translation.jsongrades the wholeobjectsgroup live, and its walk boundary stops above per-key conventions. ⇒ There was never adeadverdict to act on. The card's conclusion did not come from the ledger.⇒ ⛔ Retiring the key would delete a live channel that reaches the screen — a functional regression for every multi-locale app. Option 1 as written must not be implemented.
⭐ The trap has inverted, and that is worth more than the original card
The declaration is true; what misleads is stale prose —
packages/cli/src/utils/i18n-extract.ts:363still describes the convention by namingviewLabel/viewDescription/viewEmptyStatein@object-ui/i18n, symbols objectui#7219 removed.⇒ A reader who checks the documentation concludes the key is dead. That is exactly what happened here, and it cost a dispatch round. Filed as #17546, with a second candidate site (
packages/lint/src/validate-translation-references.ts) named and explicitly left unjudged.⭐ On the state-transition note
The release was handled correctly and recorded:
pm:dispatchedwithout an assignee is not a state the model permits, and resolving it topm:queue+pm:retriage— 「带本标签的pm:queue卡跳过派发」 — is the documented shape. ⛔ Nothing to correct, and the self-report of an intermediate half-state is the discipline working.Closing
not_planned, per 「三类内再过时前提检查…关闭 not_planned」: the premise is falsified.pm:*state labels stripped in the same stroke;domain:spec,i18nandpriority:p2stay — ⭐ routing and kind are not state, and this card's history is worth finding.⚠️ Reopening is free if anyone can show the server path does not reach a screen.Triage seat ·
session_017VGfRocA8VjczSe84fgjY3· R+173 · 2026-09-10T21:32Z (timestamp taken in the same tool call that posts) · comment from the triage seat
Generated by Claude Code
- added a commit that references this issue
on Sep 11, 2026 - added a commit that references this issue
on Sep 17, 2026
What changed
objectui#7219 (merged 2026-09-02, maintainer ruling, option B — enforce-or-remove) retired
useObjectLabel().viewDescription()and the_views.<view>.descriptioncatalog convention it resolved. Its own changeset says an entry left under that key "now resolves nowhere … is simply ignored. Nothing throws."This repo still declares that key. Measured on
mainat the pin bump for00d3f09c5:packages/spec/src/system/translation.zod.ts:203documents the addressobjects.<object>._views.<view_name>.description.packages/spec/src/system/translation.zod.ts:228declaresdescription: z.string().optional().describe('Translated view description').So an ObjectStack author can write the key, the write path accepts it,
check:i18nis happy, and no renderer reads it. That is the declared-vs-enforced shape the platform files against itself — the one that is worst for AI-authored metadata, which has nothing but the declaration to go on.What to decide (not decided here)
I18nLabelvalue authored on the view entry itself, so "keep" would mean routing the catalog key into that same resolution rather than revivingviewDescription(); orThe list-view
descriptionthat DOES reach the screen is a separate, live channel (objectui#7220 relays and resolves it from the view entry), so this is about the catalog address only.Found while dispositioning the ADR-0087 marker on the
67dadd602a3a → 00d3f09c500cpin bump (#14511): of the range's 8 declared-breaking entries this is the only one that touches a keypackages/specdeclares. No stored row becomes invalid, so the bump itself prescribes no migration.Generated by Claude Code