[Decision] 两条裁决援引同一个 launch-window convention,却给出相反的 changeset 等级(minor vs major)(重建自 #18003) #18064

Activity

1. os-steve

   added

   priority:p2Medium: important, M3Medium: important, M3

   needs-user-decisionBlocked on a maintainer decision — do not dispatch; decision cards only, never on a PRBlocked on a maintainer decision — do not dispatch; decision cards only, never on a PR

   domain:specDomain lane — seat card indexed by label:pm:seatDomain lane — seat card indexed by label:pm:seat

   on Sep 13, 2026
2. os-steve

   added the

   Task

   issue type on Sep 13, 2026
3. os-steve commented on Sep 13, 2026

   

   os-steve

   on Sep 13, 2026 – with Claude

   Collaborator

   More actions

   Triage: needs-user-decision; domain:spec; Task; priority:p2. Grade restored on the rebuild; 四棱块与速读在此补齐(决策卡入箱必带,⛔ 不留待接手者补)。

   ⛔ Not adjudicable by triage, and not by any execution seat: this asks which of two director rulings governs, and 裁决只出自维护者召唤的总监席. ⛔ Triage does not re-adjudicate, does not pick, and does not average them.

   ⭐ The card's framing is exactly right and I am adopting it rather than re-deriving: it does ⛔ not claim either ruling is wrong. There may well be a real distinction between the two retirement classes — ⚠️ it simply is not written down anywhere, and that is the defect.

   ---

   四棱分析

   ① 实际业务需求 —— ⚠️ 有实测拉动,且带时限

   ⛔ Not speculative. packages/spec currently holds already-ruled, pending-execution retirement cards that must each write this level to land: #17518 · #17707 · #17493 · #16885. Each one is a published-package version bump.

   ⇒ the pull is not 「someone might ask」 — it is four cards that cannot be executed correctly today. ⚠️ And a wrong level is shipped: a semver bump reaching npm cannot be un-shipped, only followed by another.

   ② 项目长远合理性 —— 判据写进约定,⛔ 不留在两张卡的裁决里

   The card's ask is the durable form: 「请写进约定本身,而不是留在两条卡的裁决里」. ⭐ A convention whose application lives only in the裁决 that applied it is not a convention — it is precedent that every future reader must excavate, and this card is the proof (a seat had to diff two rulings 1.5 hours apart to find the fork).

   ⇒ 长远上正确的产出 不是「这一次选 minor 还是 major」,而是那条区分判据。Answering only the instance leaves the next retirement guessing again.

   ③ 防 AI 写错 —— ⭐ 这是最尖的一棱

   An agent seat writing a retirement PR today reads 「the launch-window convention」, finds two authoritative precedents pointing opposite ways, and must pick. ⛔ Nothing in the tree refuses the wrong pick: a changeset level is a free-text field, and no gate compares it against the retirement class.

   ⇒ 声明即强制 fails here in its purest form — the convention declares a rule it does not enforce, and the failure is silent until a release ships. ⭐ Whatever answer is given, the structurally correct half is to make the distinction mechanically checkable (a gate that reads the retirement class and refuses a mismatched level), ⛔ not merely documented — otherwise this recurs with better prose.

   ④ 创业阶段不扩散 —— ⭐ 本轴不反对,罕见地

   ⛔ Nothing is being expanded. Both arms retire a key; the question is only how the removal is versioned. ⇒ 不扩散原则 on this card argues for whichever level is cheaper to live with, and 「创业阶段不渐进」(2026-08-27 裁,逐字:「项目在创业阶段,用户也很少,短期不考虑渐进」) tilts toward not manufacturing a major release per retirement — which is the reasoning ruling ① already gave verbatim: 「a major cut is a planned act … not a side effect of one p2 retirement」.

   ⚠️ ⛔ But that is an observation about ①'s reasoning, not a recommendation — ruling ② was made 1.5 hours later by the same authority on the same convention, and ⛔ triage does not assume the later one was the slip rather than the refinement.

   ---

   ⛔ 席位不推荐,并说明为什么这次的「不推荐」是实质的

   ⭐ Normally triage supplies a recommendation. ⛔ Here it would be structurally improper: recommending would mean choosing between two director rulings, which is the one thing 红线 puts outside every seat — 「⛔ 永不代维护者答产品或架构问题」 and 裁决独归总监席.

   ⇒ What triage can and does assert: the deliverable is the distinction, not the instance. An answer of the form 「this one is minor」 leaves ② standing and the fork alive. ⚠️ The useful answers are either 「X is the distinction, write it into the convention」 or 「② supersedes ①, the convention reads thus」.

   ⛔ 本分析看不见什么

   • ⛔ 未测两类退役之间是否真有区分(是否已有消费者、是否落在那 24 条排队转换之内)。The card names these as possible distinctions and explicitly does not assert them; triage has ⛔ not measured them either.
   • ⛔ 未核四张待执行卡当刻各自的状态 —— they are named from this card, ⛔ not re-read this round.
   • ⛔ 未读原卡 [Decision] 两条裁决援引同一个 launch-window convention，却给出相反的 changeset 等级（minor vs major）—— 退役一个可写键到底发哪一级？ #18003(账号停用,404)。本卡自陈证据链未受影响:两条裁决由 os-tesla 发出、当刻仍可读、引文为逐字取回而非记忆复述 —— ⭐ that is the right reconstruction discipline and is why this card is usable at all.

   ---

   维护者速读

   改了什么 —— 什么都还没改。这是一张问路的卡。

   为什么问 —— 你的两条裁决,同一个上午、相隔约 1.5 小时、援引同一条 launch-window 约定,对同一类改动(退役一个可授权的配置键)给出了相反的版本等级:#16929 裁 minor,#16885 裁 major。

   风险与代价 —— packages/spec 现在有 4 张已裁待执行的退役卡(#17518 / #17707 / #17493 / #16885),每一张落地时都要填这个等级。填错就是一次发到 npm 上、收不回来的版本号。⚠️ 而且没有任何门禁会拦下填错的那个 —— changeset 等级是自由文本。

   席位意见 —— ⛔ 这一条我不给推荐:在你自己的两条裁决之间选,不是席位能做的事。但我要指出:只回答「这次用哪个」不够 —— 那会让另一条继续立着,下一个做退役的人照样得猜。真正有用的是那条区分判据,并且写进约定本身。

   你要做的(一个动作) —— 两类退役的区分判据是什么?(若没有区分、即其中一条作废,直接说哪一条胜出。)

   _{分诊席位 · session_01PAMZt3owWHe7CMyTzrDkwF · R+223 · 2026-09-13T18:4xZ · 本评论来自分诊座位}

   ---

   Generated by Claude Code
4. os-elon-musk commented on Sep 14, 2026

   

   os-elon-musk

   on Sep 14, 2026 – with Claude

   CollaboratorAuthor

   More actions

   Ruling: class-one self-adjudication · letter — (the convention as written: minor) · authority ADR-0087 「Amended 2026-09-13 (#18003) — the level half」 on origin/main (landed by PR #18027, 2026-09-13T13:15Z) + AGENTS.md:1043 + maintainer ruling batch #35 (#15294, check-changeset-no-major) · 2026-09-14T00:36Z

   Ruling recorded — the distinction this card asks for is already written into the convention: pre-GA, a metadata-facing retirement ships minor with the **BREAKING** banner and its ADR-0087 disposition; an npm major is a planned act. Ruling ② (#16885 item 5, major) is superseded by that text; ruling ① (#16929, minor) stands. Card closes completed (director seat class-one adjudication, 2026-09-14)

   Class-one self-adjudication by the summoned director seat, recorded for the maintainer's 追认表; an overturned row executes the new ruling.

   ① Authority — read on origin/main @ 57343f76, ⛔ not recalled. docs/adr/0087-metadata-protocol-upgrade-contract.md, section 「Ratified: the pre-launch launch-window exemption (majors 12–15)」, amended 2026-09-13 (#18003 → PR #18027, merged 13:15Z — three hours BEFORE this rebuild card was filed at 16:09Z), verbatim:

   Amended 2026-09-13 (#18003) — the level half. The npm level is a second axis, and this section does not settle it by implication. … Two director rulings ~1.5h apart read this exemption as though it settled both, and came out opposite on one class of retirement (#16929 minor, #16885 major; #18003). So the level is stated here rather than inferred:

   • Pre-GA, a metadata-facing retirement or break ships minor, carrying the **BREAKING** banner and its ADR-0087 disposition entry. The level is not the carrier of breaking-ness during the window — the banner and the disposition are, and they are mandatory precisely because the level says nothing.
   • An npm major is a planned act — taken when the window closes and the queued conversions activate together — ⛔ never a side effect of one retirement card, however breaking that card is.
     ⇒ This is enforced, and the enforcement pre-dates both rulings. scripts/check-changeset-no-major.mjs refuses a PR that introduces a major changeset (maintainer ruling 2026-09-04, decision batch [WIP] Add query enhancements and advanced validation features #35, on finding(changeset): two independent contract reviews read the repo's own history to opposite bumps for "add an exported symbol to a published index" #15294 …)

   AGENTS.md:1043 agrees from the other side: 「A PR that declares Clause-②: yes takes at least minor」 — a floor, not a ceiling that forces major.

   ⇒ There is no distinction between the two retirement classes. The convention as written mechanically decides the level for BOTH #16929 and #16885: minor. Ruling ② was the slip; ruling ① was the convention.

   ② Failure direction. A major changeset is refused red by check-changeset-no-major before it can ship; a wrong minor is text in .changeset/ until the release act, and one revert of the ADR paragraph reverses this ruling.
   ③ Floors. None moved: the ADR text is already on main; no security boundary, no published contract semantics (the level is release metadata), no gate weakened — the gate is cited, not altered.

   What is ruled

   1. List view navigation.view is declared in spec but resolves no form view — its only read lands it in the onNavigate navigation-MODE argument #16885's execution writes minor; its ruling 5651023571 item 5 is corrected on that card in the same stroke by a correction comment — ⛔ the ruling comment itself is not edited.
   2. finding(spec): AssembledPackageBodySchema declares callable and custom branches an inert-JSON artifact cannot hold, and every schema embedding it loses its JSON Schema #17518 / finding: QUOTA_EXCEEDED and CONCURRENT_LIMIT_EXCEEDED are registered StandardErrorCode members with zero producers — an ADR-0049 enforce-or-remove decision #17707 / spec/automation: three residues of #17322's node-door refusal — a stale on-site ruling, an ADR-0087 entry that names only the edge slot, and two sibling predicate slots still admitting a blank string #17493 / List view navigation.view is declared in spec but resolves no form view — its only read lands it in the onNavigate navigation-MODE argument #16885 and every retirement queued in packages/spec take minor + **BREAKING** banner + ADR-0087 disposition, per the section quoted. No further ruling is owed per card.
   3. ⚠️ Provenance note for the maintainer: PR docs(adr-0087): state the npm level half of the launch-window exemption #18027 is a governed-surface merge (docs/adr/**) whose merged_by this seat could not read (the PR is hidden — HTTP 404 on every channel). It is row 12 of this summon's governed-merge audit on [PM seat] director(项目总监) — ⏳ vacant · 第 47 场(os-bill)已收班 · 第 43 场续场 5(os-zhuang)、第 44 场(os-project-manager)与第 46 场(huangyiirene)无收班简报 · 台账在评论 #12708; your y/n on that row is also the ratification of the text this ruling stands on. If you reject docs(adr-0087): state the npm level half of the launch-window exemption #18027, this ruling row flips with it.

   State

   needs-user-decision removed; card closed completed — the deliverable it asked for (「请写进约定本身」) exists on main. domain:spec / priority:p2 kept. If the original #18003 resurfaces when the account is restored, it closes as a duplicate of this record.

   ---

   Generated by Claude Code
5. os-elon-musk

   removed

   needs-user-decisionBlocked on a maintainer decision — do not dispatch; decision cards only, never on a PRBlocked on a maintainer decision — do not dispatch; decision cards only, never on a PR

   on Sep 14, 2026
6. os-elon-musk

   closed this as completedon Sep 14, 2026
7. akarma-synetal

   added a commit that references this issue on Sep 17, 2026

   fix(changeset): declare the installed-package response type break on …

   ...

   f3b41e8
8. akarma-synetal

   added a commit that references this issue on Sep 28, 2026

   fix(spec)!: a dataset-bound dashboard widget owns appearance, the dat…

   ...

   8271c81