ReviewOS

also looking at this

stacks/rfcs

RFC 0006: classify every SHOULD and MAY in Protocol 1.0 Part I

#7
Open chris wants to merge rfc-0006-should-and-may into main
1 file +174 -0
8 days ago

Closes the remaining item of stacksjs/stacks#2050: every SHOULD, SHOULD NOT and MAY in Part I is either a catalog requirement or a recorded informative statement with a reason.

  • 6 new SHOULD requirements: CORE-ROUTE-03 (resource routes, parameter constraints), STD-GEN-01 (models drive artifacts), STD-TYPE-02 and STD-TYPE-03 (inferred schema types and typed config, in typed hosts), STD-API-02 (a generated API description), STD-QUEUE-02 (delayed and recurring jobs).
  • 1 raised to MUST: CORE-REPORT-01. §16 says reports SHOULD link CI logs and the suite revision, but RFC 0004's report schema already requires execution.ci.runUrl, protocol.suiteVersion and protocol.rfcsRevision. The prose is aligned with the schema, so no conforming report changes.
  • 16 informative uses on 14 lines: permissions and definitions, two obligations on the protocol project itself, and two architecture recommendations that can't be observed from outside an implementation. Each names the requirement that carries any condition attached to it.
  • Result semantics: a SHOULD is run and reported, but never gates a profile.

The count is corrected here too. White-paper's inventory read only the first keyword per line and reported 65 uses and no SHOULD NOT. Since white-paper 086a4bf it counts all of them: 90 uses, of which 23 are SHOULD/SHOULD NOT/MAY.

This is a behavioral change under RFC 0001, so it is a Draft for maintainer review (14-day window). The catalog, schema and normative:check changes follow once it is accepted. Two questions are left open in the RFC: whether STD-GEN-01 belongs at Core, and whether MUST lines get line-level traceability here or in a follow-up.

bun run check passes.

🤖 Generated with Claude Code

1 changed file on the files tab, with 0 review threads.

Sign in to comment on this pull request.