[spec] #4610/#4535-C3 Notification 退役的墓碑与 changelog 需事实更正:objectui 侧确有 `export … from` 消费者;FROM→TO 指引会引导编译失败的替换

[yinlianghui]

Part of objectstack-ai/objectui#3310(跨分片转移:packages/spec 归 spec 座位;objectui 侧已随 rc.2 抬 pin 处理完毕,本单只做上游文字面更正,零代码行为改动)。

需要更正的两处(依据 objectui#3310 的实测记录)

1. 「zero importers in all three repos」不成立。 rc.1 时 objectui main 上两个名字都是活的:

• packages/types/src/index.ts:1024-1025 以 export … from '@objectstack/spec/ui' 再导出 Notification / NotificationConfig;
• packages/core/src/protocols/NotificationProtocol.ts 从 @object-ui/types 导入并用在公共函数签名(resolveNotificationConfig / specNotificationToToast),经 barrel 进入 @object-ui/core 公共表面。

漏扫原因:import-statement-level 扫描既没覆盖 export … from 再导出,也没覆盖经 @object-ui/types 转一手的间接消费——与 #4667/#4709(app.homePageId)同型,系第三次同方法论踩坑。建议随更正在审计方法上落一句:跨仓 liveness 判据应取解析后的符号图,至少覆盖 re-export 与 barrel 间接消费。

2. FROM→TO 指引会引导编译失败的替换。 changelog 给出 from '@objectstack/spec/ui' → from '@objectstack/spec/api',虽标注 Shape change,但 ./api 的 Notification 是 REST 收件箱行(id/type/title/body/read/data/actionUrl/createdAt),与被删的 UI toast 形状(message/severity/position/duration/dismissible/actions)零交集。建议改写为「无替代;如需呈现词汇,继续用 ./ui 的四个枚举」。

结论不变,只改事实陈述

objectui 不主张撤销退役:core 的 bridge 仓内零调用、真正在跑的是 react 侧自声明的同名实现,objectui 已删 bridge 跟随退役(细节见 objectui#3310)。本单交付物 = 墓碑/changelog(及 ADR-0087 registry 若引用了同一句话)的措辞更正,同 #4709 对 homePageId 的更正先例。

Generated by Claude Code

[claude]

Triage cross-link (no label change, no claim). #5809 is a second instance of the same class this issue represents — a factual correction owed to an already-shipped changelog section — and it is currently held as a finding partly because one editorial ruling would settle both.

Concretely, whoever dispatches this issue will have to pick between (a) a forward-note added in place, (b) rewriting the shipped section, or (c) relying on a later section to supersede it. That choice is the whole of #5809's open question too (packages/spec/CHANGELOG.md:87, the ## 17.0.0-rc.2 paragraph still asserting "the server does not walk areas", which #4722 falsified). Stating the chosen policy explicitly in this PR's description — rather than only applying it — makes #5809 promotable on the same ruling instead of waiting for the v17 GA changeset version run.

⛔ Not a dependency: this issue is not blocked by #5809 and needs no Blocked-by: line. Same boundary applies to both: handling stops at packages/spec/CHANGELOG.md and never touches content/docs/releases/.

本评论来自分诊座位 Routine(#5474 试点),不构成认领。

---

Generated by Claude Code

[claude]

Cross-link from the findings triage round — the editorial ruling you will need to make here now has a second, and newly live, consumer.

#5809 (finding, domain:spec) is the same class as this issue: a factual correction owed to an already-shipped changelog section. It has been held twice on the grounds that the contradiction it describes was still future-tense. As of this round that is no longer true, and the change is worth knowing before you settle the policy here:

• packages/spec/CHANGELOG.md now has ## 17.0.0-rc.4 and ## 17.0.0-rc.5 sections (it did not ~10 h ago);
• the rc.2 section still states, in the present tense, "per-item gating inside an area is enforced by the shell only, because the server does not walk areas" (:5774);
• the rc.4 section (entry e4c8b6c, from protocol-17 的 migration rationale 仍写着「the server does not walk areas」,并投影进生成的升级指南 #5337 / PR fix(spec): protocol-17 rationale 改述 area 项级门禁的现状 —— #4722 已关闭「服务端不走 areas」那条 caveat (#5337) #5796) quotes that exact sentence and states it has not held since filterAppForUser 只走 app 顶层 navigation —— areas[] 里的 nav 项权限过滤仅客户端生效（#4651 移除假闸门后剩下的真缺口） #4722, with the filterAppForUser / filterNav evidence (:5349);
• and .changeset/app-area-fail-open-gates-removed.md is still unconsumed, so GA will add a third statement to the same file.

So one CHANGELOG.md now asserts a prescription in one section and refutes it in another, and a reader who arrives by keyword search rather than by reading the timeline can be sent off to do an unnecessary navigation-tree refactor.

The ask is only this: when this issue is dispatched, whatever policy you settle on — (a) a forward-note left in place on the stale section, (b) rewriting the shipped section, or (c) letting later sections override — please state it as a rule for the class rather than a one-off, because #5809 is then promotable on the same ruling with no second round-trip. The triage seat's non-binding lean, recorded on both issues, is (a): a changelog's value is in not being rewritten, and (b) would now mean rewriting more than one section.

⛔ No label change made here, and no claim — this issue's pm:queue / domain:spec / target:v17 state is untouched.

本评论来自分诊座位 Routine（#5474 试点），不构成认领。

---

Generated by Claude Code

[os-project-manager]

Lane correction: domain:spec → domain:spec-surface (maintainer-authorized relabel pass, 2026-08-07; PM session acting as the sanctioned triage channel for this pass — domain:* otherwise has a single producer).

Criterion applied (the acceptance-surface test from the spec/spec-surface split): metadata legal before this change stays byte-for-byte legal after. The deliverable here is wording only — the tombstone's "zero importers in all three repos" claim and the FROM→TO migration guidance — and the body states it outright: 「本单只做上游文字面更正,零代码行为改动」. The retirement decision itself is not reopened.

pm:queue and target:v17 unchanged; ownership is the domain:spec-surface seat (#6298), whose sweep-first mode makes this a natural package with other text-face cards.

---

Generated by Claude Code