[JarryShaw]

Please follow the guide below

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort) -- no pcapkit/ file is touched
• make test passes, and a test case covers the change -- the page is pinned in test_conventions_doc_claims.py's ANCHORS
• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible -- N/A, centralised in chore(docs): centralize 1.6.0 changelog updates in one place, not per-feature-PR #2449

What is the purpose of your pull request?

• docs — documentation only

---

Description of your pull request and other information

Adds contributing/conventions/protocol-family-grouping.rst: the id() criterion, abstract family names, grouping by name, selector matching, canonical names and the _Family rule, each with its ruling's source. Sphinx adds no new warnings. Closes #2476

[JarryShaw]

NEEDS CHANGES (Sonnet, 3a4651589). Every code fact the page cites is accurate, and the docs build adds no warnings. Four statements go beyond the rulings:

• "Lists its direct concrete subclasses, and nothing else" is false on main: VLAN, L2TP and HTTP list their own name first (measured). The 16:08Z ruling decides which subclasses are listed. Scope the rule to that.
• "One consequence is accepted" is my own caveat, since declared moot (A malformed known IPv6 extension header should parse as Raw, not IPv6_Ext #2477). Drop it and the Shim6 narration.
• The SSL/TLS/DTLS grouping names nothing in any release tag and isn't confirmed yet. Cut it; HTTP, IP and IPsec carry the rule.
• The flagged inference line: the ruling only says a versioned class name "should be fine". Use "may" or delete it.
• Also: use :pr: for feat(application): implement TLS/SSL and DTLS record layer + handshake framing #2441/feat(internet): implement ICMP and ICMPv6 under an abstract ICMP base #2446; drop "Three shapes"; scope the selector claims (ProtoChain also matches aliases); add RARP/DRARP to the ARP row.

[JarryShaw]

Verified (80a11f953): the page now states only the rulings plus measured main facts; "nothing else", the SSL sentence and the Shim6 caveat are gone; :pr: roles fixed. Sonnet review dispatched.

[JarryShaw]

NEEDS CHANGES (Sonnet, 80a11f953). The rulings are now faithful, the measured facts hold, and docs warnings equal main's. Remaining:

• HTTP example misattributes the cause. 'HTTPv2' matches an HTTP/1.1 frame because the parsed layer is the HTTP dispatcher, whose id() lists both. Narrowing HTTPv1/HTTPv2.id() doesn't change that. Illustrate the family-inclusive shape with VLAN ('S_Tag' in a C-Tag frame) instead.
• Only RARP is bound to 0x8035. DRARP is an alias of a parsed RARP.
• Drop the inferences: "pick a shape by reach", "InARP exists for explicit construction", "the tuples are settled here".
• Canonical names: state both rulings with their sources. A versioned protocol's display name carries its version (feat(internet): add IGMPv1/IGMPv2 as concrete classes over an abstract base #2444, 14:27Z). ICMPv4 takes the RFC 792 name, which has none (feat(internet): implement ICMP and ICMPv6 under an abstract ICMP base #2446, 19:13Z).
• Nits: the _Family source is feat(internet): implement ICMP and ICMPv6 under an abstract ICMP base #2446 at 17:14Z; define "registry key"; replace "the second bullet below" with a named reference.

[JarryShaw]

Verified (71e86243c): HTTP example replaced by VLAN (measured); DRARP fixed; inferences dropped; both canonical-name rulings stated with sources. Sonnet review dispatched.

[JarryShaw]

NEEDS CHANGES (Sonnet, 71e86243c). Verified facts and the Sphinx build hold. Three fixes:

• Stale on main (now merged): "On main, IPv6_Ext.id() is ('IPv6_Ext',)" is false since fix(protocols): compose IPv6_Ext.id() from its direct concrete members #2464 merged; it is the eight-member tuple. Drop the sentence, and the "checked against main" body line with it.
• Stronger than sources: feat(internet): implement ICMP and ICMPv6 under an abstract ICMP base #2446's ICMPv4 naming is "should probably drop", not "rules". feat(internet): add IGMPv1/IGMPv2 as concrete classes over an abstract base #2444's "I'd prefer" is a preference, not a flat rule.
• Timed context: the (14:27Z), (19:13Z), 17:14Z tags; cite :pr: alone.

[github-actions]

Coverage: 94.44% (unit tier, Python 3.15, ebd7730f6, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19609	133	2620	120	98.84%
pcapkit/corekit	2605	74	818	37	96.52%
pcapkit/dumpkit	281	3	100	3	98.43%
pcapkit/foundation	3070	24	1072	26	98.79%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17614	193	4678	179	98.32%
pcapkit/toolkit	924	36	334	10	95.55%
pcapkit/utilities	502	7	144	5	98.14%
pcapkit/vendor	4449	1854	1020	167	55.11%

Per-file detail: the coverage-html artifact of this run.

[JarryShaw]

NEEDS CHANGES (Sonnet, 509c5f694). id() tuples and selector claims hold on main.

• Member Shapes: "selected only by its own name and its family's key" is false for AH/ESP; IPv6_Ext is selected by each family it belongs to.
• Canonical Names: attribute the feat(internet): implement ICMP and ICMPv6 under an abstract ICMP base #2446 proposal to the owner's reply, as extension-header-subclassing.rst does; drop the timed "until ICMP lands".
• id() Criterion: drop "since that intermediate carries its own tuple" (not in the cited feat(application): implement TLS/SSL and DTLS record layer + handshake framing #2441 comment).
• Non-blocking: sibling-selection sentence repeats at L94-97 and L124-126.

[JarryShaw]

NEEDS CHANGES (Sonnet, 2521ac182). Measured on main (4ca3f30d7, has #2464): the IPv6_Ext key selects AH and ESP. The previous fix's claim held only for the pre-#2464 base.

• Member Shapes (~L96-98): delete "The IPv6_Ext key does not select them…"; say AH and ESP are selected by IPsec and by IPv6_Ext.
• ~L106 "On main," is a snapshot qualifier; drop it. PR body "checked against main" line: drop.
• Non-blocking, owner call: :pr: cited seven times; documentation.rst says cite the issue.