Filed unassigned while discharging #10229 (PR #10256). Observation, not a defect with a prescribed remedy — no pm:queue.
What was measured
#10229 reported three stale prose version claims. Sweeping the two shapes across the repo found five, in three files, and the reason the count grew is that the same prose is written out more than once.
The tag-scheme sentence exists three times, in three files, all saying the same thing:
content/docs/deployment/self-hosting.mdx:59 `@objectstack/cli` versions (`17.0.0`, `17.0`, `17`, `latest`) and the image is
content/docs/deployment/index.mdx:42 whose tags mirror `@objectstack/cli` versions (`17.0.0`, `17.0`, `17`,
content/docs/upgrading.mdx:36 `@objectstack/cli` versions (`17.0.0`, `17.0`, `17`, `latest`). **Pin the exact
The "two clocks" table exists twice — content/docs/deployment/index.mdx:10-17 and content/docs/upgrading.mdx:8-14 — with three rows byte-identical (Ships as, Versioned by, Whose cadence) and the remainder deliberately different (The deploy action vs The upgrade action, and so on). The Versioned by row carried the same stale our release train (17.0.0) in both.
So one publish produced five separate repairs, of which five out of five were one of two sentences. The duplication is what turned a two-sentence drift into a five-site drift.
Why this is only an observation
The pages are not redundant — a reader arriving at "how do I upgrade" and one arriving at "what am I deploying" both need the versioning frame, and the two tables genuinely diverge after row three. Extracting a shared MDX partial is one option; accepting the duplication and guarding it is another, and #10229 took the second: PR #10256 enumerates all five claim sites in check-docs-image-tag.mjs's new PROSE_CLAIMS limb, so a literal reintroduced into any copy is now caught.
That means the drift is guarded either way and this card is not urgent. What it records is the standing cost: every future edit to the versioning frame has to be made in two or three places, and the gate enforces consistency of shape (no concrete version) rather than consistency of content. If the pages ever disagree about something the gate cannot see — the way self-hosting.mdx:60 was false while its siblings were merely stale — nothing catches it.
Not claimed
Found from #10229 / PR #10256.
Generated by Claude Code
Filed unassigned while discharging #10229 (PR #10256). Observation, not a defect with a prescribed remedy — no
pm:queue.What was measured
#10229 reported three stale prose version claims. Sweeping the two shapes across the repo found five, in three files, and the reason the count grew is that the same prose is written out more than once.
The tag-scheme sentence exists three times, in three files, all saying the same thing:
The "two clocks" table exists twice —
content/docs/deployment/index.mdx:10-17andcontent/docs/upgrading.mdx:8-14— with three rows byte-identical (Ships as,Versioned by,Whose cadence) and the remainder deliberately different (The deploy actionvsThe upgrade action, and so on). TheVersioned byrow carried the same staleour release train (17.0.0)in both.So one publish produced five separate repairs, of which five out of five were one of two sentences. The duplication is what turned a two-sentence drift into a five-site drift.
Why this is only an observation
The pages are not redundant — a reader arriving at "how do I upgrade" and one arriving at "what am I deploying" both need the versioning frame, and the two tables genuinely diverge after row three. Extracting a shared MDX partial is one option; accepting the duplication and guarding it is another, and #10229 took the second: PR #10256 enumerates all five claim sites in
check-docs-image-tag.mjs's newPROSE_CLAIMSlimb, so a literal reintroduced into any copy is now caught.That means the drift is guarded either way and this card is not urgent. What it records is the standing cost: every future edit to the versioning frame has to be made in two or three places, and the gate enforces consistency of shape (no concrete version) rather than consistency of content. If the pages ever disagree about something the gate cannot see — the way
self-hosting.mdx:60was false while its siblings were merely stale — nothing catches it.Not claimed
versions (X.Y.Z, ...)andrelease train (X.Y.Z)). Other duplicated deployment prose was not surveyed.Found from #10229 / PR #10256.
Generated by Claude Code