You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
docs(nav): releases sits in the Platform group but every release page declares itself written for app developers #8966
Restart-when: any PR touches the root content/docs/meta.json group layout
Observation recorded while implementing #8909 (see PR #8964). Not fixed there — that card's file surface is the new upgrading.mdx plus its own nav line, and this is a separate structural decision.
The mismatch
Root content/docs/meta.json places releases in the Platform group, beside kernel and protocol:
But content/docs/releases/index.mdx states its own audience explicitly:
Each release page is written for third-party developers building apps, plugins, or clients on ObjectStack. It leads with breaking changes and migration steps...
And the content bears that out — the v17 upgrade checklist is 66 bullets about allowExport, sharedWith.type, RLS priority, and agent.tools[], all metadata-app authoring concerns, none of them kernel or protocol internals.
So the pages are shelved next to the two most internals-facing sections in the docs while being written for the audience that lives in Build.
Why it is only an observation
#8909's card names this mismatch and resolves it a different way — by putting the new upgrading.mdx door in Build rather than by moving releases. That door now exists and links every per-major checklist, so the acute discoverability problem is closed. What remains is the placement itself, which is a maintainer call about nav taxonomy, not an implementation detail:
Leave it — release notes are arguably a platform-train artifact regardless of who reads them, and upgrading.mdx now provides the audience-appropriate entry point. Zero churn.
Move releases into Build — matches the stated audience, but puts a growing per-version archive in the group app developers browse daily, and moves a section every existing inbound link points at.
No behaviour is broken either way; this is discoverability and taxonomy. Recording it so the deployment-docs restructure round can grade it rather than having it noticed a fourth time. Note #8913 re-orders content/docs/deployment/meta.json, a different file, so it does not cover this.
Triage first-touch grading (2026-08-16, triage seat Routine): hold — finding → pm:on-hold, type Task.
Reason. The acute discoverability problem this records is already closed by #8909's upgrading.mdx door (the card says so itself), so there is no measured business pull today: no reader is stranded, and the remaining question is pure nav taxonomy. Along the standing evaluation axes: zero pull ⇒ defer; moving releases breaks every inbound link into a section for zero functional gain (startup scope discipline says the zero-churn option wins by default); AI error-resistance is unaffected either way. This does not meet the decision-inbox bar — the options differ in shelving, not in product semantics or contract shape — so it is graded here rather than escalated.
Named restart conditions (any one re-opens the question; ⛔ no dispatch before then):
A PR touches the root content/docs/meta.json group layout (the trigger file for this card — domain:devx seat: pre-dispatch check for cards touching that file should name this card and treat the move as a declarable rider, per the opportunistic-restart discipline);
The consumed trigger is retired so it cannot re-fire. Remaining restart conditions: ① any card touching root content/docs/meta.json (devx seat: keep it in the pre-dispatch must-check list — that card takes this as a declared rider or declines on-card); ② a maintainer statement on releases-section placement.
Triage: this comment comes from the triage seat Routine; not a claim.
Activity
os-project-manager commentedon Aug 16, 2026
Triage first-touch grading (2026-08-16, triage seat Routine): hold —
finding→pm:on-hold, typeTask.Reason. The acute discoverability problem this records is already closed by #8909's
upgrading.mdxdoor (the card says so itself), so there is no measured business pull today: no reader is stranded, and the remaining question is pure nav taxonomy. Along the standing evaluation axes: zero pull ⇒ defer; movingreleasesbreaks every inbound link into a section for zero functional gain (startup scope discipline says the zero-churn option wins by default); AI error-resistance is unaffected either way. This does not meet the decision-inbox bar — the options differ in shelving, not in product semantics or contract shape — so it is graded here rather than escalated.Named restart conditions (any one re-opens the question; ⛔ no dispatch before then):
content/docs/meta.jsongroup layout (the trigger file for this card —domain:devxseat: pre-dispatch check for cards touching that file should name this card and treat the move as a declarable rider, per the opportunistic-restart discipline);meta.json— lands last #8913 (the deployment-section restructure keystone) lands — its author is instructed to describe where content ended up, which is the natural moment to grade thereleasesplacement once, with the whole restructured tree in view;Provenance: seat-level grading by the whole-repo triage Routine (first-touch finding round), not a maintainer ruling.
Generated by Claude Code
os-project-manager commentedon Aug 16, 2026
Hold maintenance (triage seat, 2026-08-16): restart condition "#8913 lands" fired — #8913 closed 05:12Z. Re-checked and the hold stands:
origin/main: rootcontent/docs/meta.jsonstill shelvesreleasesunder---Platform---; docs(deployment): rewrite the section index around the two lifecycles and re-ordermeta.json— lands last #8913's change wascontent/docs/deployment/meta.json, exactly as this card predicted ("a different file, so it does not cover this").content/docs/upgrading.mdxin the Build group — state both upgrade halves and give the per-major checklists a door #8909'supgrading.mdxdoor, pull remains zero, and the residual question is the maintainer taxonomy call the card itself describes — with "current post-docs: newcontent/docs/upgrading.mdxin the Build group — state both upgrade halves and give the per-major checklists a door #8909 state by default" as a legitimate standing answer.The consumed trigger is retired so it cannot re-fire. Remaining restart conditions: ① any card touching root
content/docs/meta.json(devx seat: keep it in the pre-dispatch must-check list — that card takes this as a declared rider or declines on-card); ② a maintainer statement on releases-section placement.Triage: this comment comes from the triage seat Routine; not a claim.
Generated by Claude Code
objectstack-fleet commentedon Sep 23, 2026
关
not_planned—— 维护者逐张复核 on-hold 卡时同意关闭;保持现状就是答案分诊席(
session_01Tw7jnJinGHvoGSi8aFkhPJ),2026-09-23T09:01Z。维护者 2026-09-23 在分诊会话里逐张复核pm:on-hold卡,对第四组的回复原文:「第四组 同意」。 本卡在那一组里的建议是关闭。「发版说明」栏目放在 Platform 分组下,但它是写给应用开发者的。真正的问题(应用开发者找不到升级步骤)已经由 #8909 在 Build 分组里加的
upgrading.mdx入口解决了。剩下的只是分类问题;挪动栏目会让所有指向它的链接失效,却换不来任何功能上的好处。2026-08-16T03:15Z 分诊定级(5305481215)已经说明,「保持现状」本身就是一个正当的答案。5305481215、5306101166,都是 2026-08-16 的);卡片最后更新时间是 2026-08-20。⛔ 本席没有读到可能存在的第三条,所以不对它做任何断言(同一天已经看到过评论计数与实际不符的情况)。重开条件
维护者对这个栏目的位置有了明确偏好;或者有应用开发者反馈找不到发版说明。
关闭理由:
not_planned,同时摘掉pm:on-hold。Generated by Claude Code