Repository navigation
Open
Description
A review of #75 found five non-blocking gaps in the classification. The split itself is correct: 49 scenarios, 13 draft across 2.1/2.6/2.7, 36 library, and the embedded gherkin matches the feature file at head. These are the guardrails and wording issues left after the merge.
- The
@draftset asserts the library's section URI. Five of the seven 2.7 scenarios assertthe response body should contain "section-2.x". The draft's own 2.7 examples usetype: https://developer.example.com/idempotency, andsection-2.appears nowhere in draft-07. An implementation that meets 2.7 with its own problem type fails the draft selection. Cheapest fix: say in the 2.1 and 2.7 Rule prose that thesection-2.xtype token is this library's convention while status, content type and the presence of type/title/detail come from the draft. Moving the assertions into@librarychanges assertions, which docs(spec): classify scenarios as draft requirements or library policy #75 declared out of scope. - The gherkin in
docs/specs/index.mdis hand-synced. The fence duplicates the whole feature file and website CI runs only lint, commitlint, prettier and link checks. Add a CI step that extracts the fence body and diffs it againstdocs/specs/idempotency.feature. - The overview counts are hand-computed prose. The
13 of the 49and per-Rule bullets are not checked anywhere. Parse the feature file's Rule tags in a website check and assert the figures, or generate the bullets at build time. - The tag structure is never exercised where it is authored. Website CI does not parse or dry-run the feature file, so
--tags @draftselecting 13 is only proven in idempot-js after a sync. Run the classification checker or a cucumber dry-run in website CI. - The
@draftprologues say "requires" where the draft says SHOULD. Lines 28 and 67 read "The draft requires ..." while draft-07 uses "The resource SHOULD ...". Line 17's 2.1 MUST wording is accurate. Change "requires" to "says the resource should" in the 2.6 and 2.7 prologues.
Source review: pi-review on #75 (correctness, testing, adversarial, plus an independent validator). Non-blocking for the merge, so tracked here rather than reopening #75.
Metadata
Metadata
Assignees
Labels
No labels
Activity