Repository navigation
Convention page: protocol family grouping and the id() criterion #2476
Description
Convention page: protocol family grouping and the id() criterion
Summary
On #2441 (2026-10-30 16:20Z) the maintainer asked for the id() criterion and the subclassing/grouping logic to be saved as a convention. No page exists yet. The rules are currently spread across PR threads (#2441, #2443, #2444, #2446, #2462, #2464), and each new protocol family has re-derived them, sometimes wrongly.
Scope
A new page under docs/source/contributing/conventions/, a sibling of extension-header-subclassing.rst, covering:
- The
id()criterion (feat(application): implement TLS/SSL and DTLS record layer + handshake framing #2441, 16:08Z): a family base lists only its direct concrete subclasses, and excludes abstract intermediates and their subclasses.IPgives('IPv4', 'IPv6'), without AH/ESP, which belong toIPsec. - Grouping by name and history, not wire format.
SSLcovers the SSL versions, TLS and DTLS, asHTTPgroups HTTP/1–2. No dispatcher logic is needed where the precedent (IP/IPsec) has none. - The member shapes in use. Narrow-to-self (
AH,ESP); family-inclusive (C_Tag/S_Tag,HTTPv1/HTTPv2,L2TPv2); the ARP/InARP single class with a dynamic alias and a fixedid(), plus a thin explicit-construction subclass; a base doubling as its own default member when names collide (L2TP/L2TPv2). - What selectors actually match. Registry-key strings and class objects expand through
id(), while non-key labels match by name. Measured on feat(link): implement LINUX_SLL and LINUX_SLL2 dissectors #2443:'HTTPv2' inan HTTP/1.1 frame isTrue. - No invented suffixes. No
_Familybase exists, and none should be minted.
Blocked on
#2462's open question: whether a concrete fallback parser (IPv6_Ext) carries a family tuple or none. #2446's ICMP raises the same concrete-base question. The page should state the answer rather than guess it. Unblocks when #2462's needs: decision is answered.
Activity
JarryShaw commentedon Oct 30, 2026
Unblocked: #2462 is answered. Held only on Actions queue depth (30) while the 1.5.1 release runs.
JarryShaw commentedon Oct 30, 2026
Still held: Actions queue at 41 while the 1.5.1 release Unit Tests runs (14/32 jobs done).
JarryShaw commentedon Oct 30, 2026
Still held: Actions queue at 36 while the 1.5.1 release Unit Tests finishes (18/32 jobs done).
JarryShaw commentedon Oct 30, 2026
Still held: Actions queue at 29 while the 1.5.1 release Unit Tests finishes (31/46 jobs done).
JarryShaw commentedon Oct 30, 2026
Still held: Actions queue at 37;
main's release Unit Tests is waiting on runners for its ordering legs. No new pushes from me until it completes.JarryShaw commentedon Oct 30, 2026
Still held:
main's release Unit Tests has 14 jobs waiting for runners (Actions queue at 31).JarryShaw commentedon Oct 30, 2026
Still held: the 1.5.1 release Unit Tests re-run is in progress; no pushes until it completes.
JarryShaw commentedon Oct 30, 2026
Still held: release Unit Tests re-run in progress.
6 remaining items
JarryShaw commentedon Oct 30, 2026
Still held: 1.5.1 is on PyPI; conda builds finishing.
JarryShaw commentedon Oct 30, 2026
Still held: 1.5.1 conda builds finishing.
JarryShaw commentedon Oct 30, 2026
Still held: last 1.5.1 conda builds finishing.
JarryShaw commentedon Oct 30, 2026
Still held: one 1.5.1 conda build left.
JarryShaw commentedon Oct 30, 2026
Dispatched now that 1.5.1 is published.
docs(conventions): scope the family grouping page to the rulings (#2476)
docs(conventions): add the protocol family grouping page (#2478)