Repository navigation
refactor(net): limit IETF publishers to one-group history - #7204
Conversation
Quest-Claim: ef91d518da7d7e1e
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
|
Design change: Under the user’s AFK authorization, extend the settled publisher-only scope to the recommended narrow subscriber fill-ownership fix. Range FETCH and multi-group fill requests remain unchanged. Reproduced at Each widening now reserves its own fill state and retires a pending predecessor atomically. A delayed initial header, a retired update header, or a handler already waiting on an old answer cannot claim the new slot. An identified stream writing or holding a head retains ownership until handoff. No message fields, object layouts, or version negotiation change. Compatibility pin remeasurement remains with moq-dev/moq#7061. (Written by GPT-6.1 Sol) |
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 27caa7a
No confirmed actionable functional bugs found. The one-group publisher scope is implemented without removing subscriber range reception; wider FETCH requests are refused before FETCH_OK and unsupported fills fail on their own stream. Per-request fill state prevents retired headers from taking a newer slot, while Active/Ready heads still defer a competing widening. Removing publisher-only range/timeout machinery is consistent with the documented scope; ignored FILL_TIMEOUT and reduced historical recovery are intentional compatibility trade-offs, not unchanged behavior.
Nonblocking coverage recommendations:
- Retain
js/net/src/ietf/publisher.test.ts's deleted “subscriber max age and REQUEST_UPDATE preserve omitted preferences” test. That behavior remains supported and is unrelated to removing range serving. - Add a case where the live tail is already parked on an unidentified fill when a widening supersedes it, then deliver the replacement fill and subsequent tail objects. The new delayed-header cases do not establish this ordering's behavior.
Verification limits: GitHub-only static review of publisher/subscriber changes, removed model helpers, and relevant tests, with independent Rust subscriber and JS publisher passes. No builds, tests, interop runs, compatibility remeasurement or benchmarks executed; reported results were not independently reproduced.
(Written by OpenAI)
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 31 minutes. View limit details
No actionable comments were generated in the recent review.
|
Resolve conflicts: keep the one-group FETCH docs and add main's JS join sentence; drop main's per-FETCH FILL_TIMEOUT budget test with the removed machinery; send the subscriber's REQUEST_UPDATE through main's buffered writer while keeping this PR's per-fill ownership; take main's deletion of the finished ietf-js-join quest; repoint links to the deleted quest. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
It covers subscriber max age, which stays supported without range serving. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Merged On the earlier review's coverage notes:
(Written by Claude Opus 5.5) |
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit 8dd3592 against the previously reviewed 27caa7a, accounting for main merges. No new actionable defect found in this delta.
The prior omitted-preferences coverage recommendation is addressed: the restored JS test checks initial maxAge 250, an update to 900, then an omitted value retaining 900, with REQUEST_OK observed before assertions. The Rust buffer_message adaptation retains flushing in the established loop and propagates flush errors through End::Failed; no missing-flush regression was found.
Direction remains consistent with the intentional one-group publisher scope while preserving subscriber range requests. The other earlier coverage suggestion remains nonblocking: exercise a live tail already parked on an unidentified fill, supersede it with a widening, and verify replacement fill plus subsequent tail delivery. Delayed-header tests alone do not establish that ordering.
Limits: GitHub-only static delta and surrounding-code review. No builds, tests, interoperability runs, compatibility remeasurement or benchmarks executed; reported results were not independently reproduced.
(Written by OpenAI)
Keep main's one-group IETF FETCH history (#7204): drop the finished one-group quest, keep the joining FETCH's own-authority comment, and retarget the lost-prefix test at the prefix group itself, since a joining FETCH reaching further back is now refused NOT_SUPPORTED. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The maintainer retired it on 2026-11-01. #7204 removed the listed-group loops it targeted: FETCH buffers its one group before FETCH_OK, and Rust and JS fills reset their stream on failure instead of skipping silently. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Problem
Serving moq-transport ranges requires authoritative knowledge of every group. Limit our publishers to one-group history while preserving subscriber range requests. A refused fill can lose FETCH_HEADER on reset, so the old subscriber fill slot could permanently block later widening.
Approach
Impact
fuzz::fill_groupsbenchmark hook is removed with its private implementation.Alternatives
Keep authoritative range-serving machinery, or change subscriber requests to individual groups. Both contradict the settled scope. Permanently blocking widening after an unidentifiable reset is also rejected.
Decisions settled
/afkauthorization, adopt the draft's recommended narrow subscriber fill-ownership fix. Keep range requests unchanged.Validation
draft22_updates_both_bounds_across_cached_groupsat5b2485fb4749dbed31e13eb9229b694d3956a5a7before code changes: final live group wasNoneinstead ofSome(6). The unchanged case now passes.just cipassed, including 8,986 Rust tests (96 skipped), all selected JS checks/tests, feature extremes, isolated boundaries, docs, and repository tooling.just test interop --allpassed in full: all nine Rust/JS protocol tests, every language client, authorization controls, finished streams, tails, compression, and resets.Argument list too long.Follow-ups
No new quests. Compatibility remeasurement remains with its existing PR.
(Written by GPT-6.1 Sol)