Repository navigation
docs(conventions): Method._missing_ folds case against RFC 9110 §9.1, and registry-protocol.rst records neither the fold nor the get/constructor split #2552
Description
From the 2026-11-01 conventions audit, re-verified on 73add05a3. registry-protocol.rst, "RFC-Directed Case Sensitivity": a case fold has to cite the RFC or IANA registry that makes the values case-insensitive, else it is a defect. The Method row says get does not fold (#896); the note on a fold outside any get names only esp._resolve.
Method._missing_inpcapkit/const/http/method.pyupper-casesvalueand matches member names, then values:Method('get') is Method.GET, whileMethod.get('get')returns an unregistered member namedGETwith value'get'. RFC 9110 §9.1 makes the method token case-sensitive.- Its docstring calls the split deliberate and cites fix(const): Method.get is case-insensitive, but RFC 9110 makes the method token case-sensitive #896, but fix(const): Method.get is case-insensitive, but RFC 9110 makes the method token case-sensitive #896's thread concerns
getalone (the fix removed the fold fromget); fix(const): Method('baseline-control') mints an unregistered member, not BASELINE_CONTROL #2284 and fix(const): find hyphenated HTTP methods by value in Method._missing_ (#2284) #2351 later widened the constructor path. The rule recorded on design: should the non-registry helper enums share a lookup interface, or be made strictly immutable? #877 lets a registry fold only where an RFC recognises it. - Stale wording on the same page:
Method.get"catches it in order to mint". It returns_unregistered_member(...), which grows neither lookup table, and mint-criterion.rst defines minting asextend_enum. - Probe:
git show origin/main:pcapkit/const/http/method.py | grep -n "value.upper()".
Decision needed: does the constructor fold stay?
- Keep it: record the get/constructor split on the page with its basis, and correct the docstring's attribution.
- Drop it: remove the fold from the
_missing_template inpcapkit/vendor/http/method.pyand regenerate (observable:Method('get')stops resolving toGET).
Recommendation: 1, because fix(const): Method('baseline-control') mints an unregistered member, not BASELINE_CONTROL #2284 asked forMethod('baseline-control')to resolve toBASELINE_CONTROLand fix(const): find hyphenated HTTP methods by value in Method._missing_ (#2284) #2351 did it through this fold; dropping it reverses that.
Fix: reword "to mint" either way; option 1 edits registry-protocol.rst, which feat(registry): foundation.registry wrappers for IPX and SCTP sub-maps; record the convention (#2511) #2521 also edits.
Activity
JarryShaw commentedon Nov 1, 2026
Take 1. I remember Method should be case-insensitive. And we have to replace dashes with underscores for Python's naming conventions.
JarryShaw commentedon Nov 1, 2026
Owner chose 1, paraphrased:
Method's constructor fold stays.Method('baseline-control')isBASELINE_CONTROL, per fix(const): Method('baseline-control') mints an unregistered member, not BASELINE_CONTROL #2284 and fix(const): find hyphenated HTTP methods by value in Method._missing_ (#2284) #2351).Blocked: this edits registry-protocol.rst, which #2521 also edits, so it unblocks when #2521 merges. It then joins #2551, #2553, #2556 and #2565 in one worker.
JarryShaw commentedon Nov 1, 2026
Unblocked: #2521 merged, so registry-protocol.rst is free. Dispatched together with the page's other queued fixes (#2551, #2552, #2553, #2556, #2565) as one worker and one PR.