From 9bf0eac0c42d9ce1a5fc2c669129471c9752e7a0 Mon Sep 17 00:00:00 2001 From: Luke Curley Date: Wed, 7 Oct 2026 18:51:20 -0700 Subject: [PATCH 1/2] quest: keep peer-asked PUBLISH_NAMESPACE retries; note async credit backends Co-Authored-By: Claude Opus 5.5 --- quest/m1/js-request-deadline.md | 9 +++++++-- quest/m1/rs-request-credit.md | 5 ++++- 2 files changed, 11 insertions(+), 3 deletions(-) diff --git a/quest/m1/js-request-deadline.md b/quest/m1/js-request-deadline.md index c84970040b7..73fe2c881ea 100644 --- a/quest/m1/js-request-deadline.md +++ b/quest/m1/js-request-deadline.md @@ -34,8 +34,13 @@ Decided 2026-10-07, re-planning the external opt out of it, and their setup timer (`SUBSCRIBE_SETUP_TIMEOUT_MS` in lite, `SUBSCRIBE_OK_TIMEOUT_MS` in IETF, PUBLISH_NAMESPACE's own) covers the create as well as the answer. -- PUBLISH_NAMESPACE is in scope with no retry: drop the IETF publisher's loop - that re-offers a failed advertisement. +- PUBLISH_NAMESPACE is in scope with no retry: the IETF publisher's loop + (`runPublishNamespaces`, `js/net/src/ietf/publisher.ts`) stops re-offering + after either failure above. It still re-offers when the peer's + REQUEST_ERROR names a retry interval, and keeps its backoff for a + draft-14/15 refusal, which can't carry one (decided 2026-10-07 from review: + the peer asked for those, and dropping them leaves a namespace + undiscoverable until reconnect). - The credit error is local, with no wire code. Propose its name in the PR; it is a new public `@moq/net` error. - `@moq/watch` keeps ending a track on either failure. Rejected: #4999's diff --git a/quest/m1/rs-request-credit.md b/quest/m1/rs-request-credit.md index 4f3f7ab1b4e..1307881058c 100644 --- a/quest/m1/rs-request-credit.md +++ b/quest/m1/rs-request-credit.md @@ -24,7 +24,10 @@ which can be pending under send-window pressure. Check credit at the QUIC level instead, likely a non-waiting open on the transport trait. Also note that `moq-tokio`'s `poll_open_bi` parks an abandoned open in the session's slot rather than dropping it, so the next request inherits it; the non-waiting open -should avoid that. +should avoid that. The async backends `moq-tokio` wraps (qmux, iroh, noq) have +no credit query in `web-transport-trait`; prefer adding one upstream in +moq-dev/web-transport over leaving them waiting, and say in the PR which +backends fail fast. Propose the error variant's name in the PR. Public API: one new `Error` variant, and possibly a transport-trait method. Wire: none. From f731a235d61214cbb0754956f286fa5cc86466ce Mon Sep 17 00:00:00 2001 From: Luke Curley Date: Wed, 7 Oct 2026 20:21:36 -0700 Subject: [PATCH 2/2] quest: name both advertise loops, the 10 s advertise timer, and the poll trait From the Grok review of 9bf0eac0. Co-Authored-By: Claude Opus 5.5 --- quest/m1/js-request-deadline.md | 18 +++++++++++------- quest/m1/rs-request-credit.md | 10 ++++++---- 2 files changed, 17 insertions(+), 11 deletions(-) diff --git a/quest/m1/js-request-deadline.md b/quest/m1/js-request-deadline.md index 73fe2c881ea..9a7401300aa 100644 --- a/quest/m1/js-request-deadline.md +++ b/quest/m1/js-request-deadline.md @@ -32,15 +32,19 @@ Decided 2026-10-07, re-planning the external `js/net/src/stream.ts`) stays the default for `Stream.open` and `Writer.open`, which SETUP, probe, and group streams inherit. Request sites opt out of it, and their setup timer (`SUBSCRIBE_SETUP_TIMEOUT_MS` in lite, - `SUBSCRIBE_OK_TIMEOUT_MS` in IETF, PUBLISH_NAMESPACE's own) covers the + `SUBSCRIBE_OK_TIMEOUT_MS` in IETF, and PUBLISH_NAMESPACE's + `ADVERTISE_TIMEOUT_MS`, which moves from 5 s to the shared 10 s) covers the create as well as the answer. -- PUBLISH_NAMESPACE is in scope with no retry: the IETF publisher's loop - (`runPublishNamespaces`, `js/net/src/ietf/publisher.ts`) stops re-offering - after either failure above. It still re-offers when the peer's - REQUEST_ERROR names a retry interval, and keeps its backoff for a +- PUBLISH_NAMESPACE is in scope with no retry: both loops that advertise + through `#advertise` (`runPublishNamespaces` and the draft-14/15 solicited + `#runSubscribeNamespace`, `js/net/src/ietf/publisher.ts`) stop re-offering + after either failure above. Both failures return `"dropped"` today, the + answer that keeps the backoff running, so they need their own terminal + answer. The namespace then stays unadvertised through that peer until it is + republished or the session reconnects. Each loop still re-offers when the + peer's REQUEST_ERROR names a retry interval, and keeps its backoff for a draft-14/15 refusal, which can't carry one (decided 2026-10-07 from review: - the peer asked for those, and dropping them leaves a namespace - undiscoverable until reconnect). + the peer asked for those). Test both loops. - The credit error is local, with no wire code. Propose its name in the PR; it is a new public `@moq/net` error. - `@moq/watch` keeps ending a track on either failure. Rejected: #4999's diff --git a/quest/m1/rs-request-credit.md b/quest/m1/rs-request-credit.md index 1307881058c..fe64a214400 100644 --- a/quest/m1/rs-request-credit.md +++ b/quest/m1/rs-request-credit.md @@ -19,15 +19,17 @@ leaves, or the session closes, and a relay shouldn't kill a subscribe on a slow upstream hop. A single `Pending` poll of `open_bi` is not "no credit": on WebTransport, -`web-transport-quinn` opens the QUIC stream and then writes the session header, +`web-transport-moq` opens the QUIC stream and then writes the session header, which can be pending under send-window pressure. Check credit at the QUIC -level instead, likely a non-waiting open on the transport trait. Also note that +level instead, likely a non-waiting open on moq-net's `transport::poll` +session trait. Also note that `moq-tokio`'s `poll_open_bi` parks an abandoned open in the session's slot rather than dropping it, so the next request inherits it; the non-waiting open should avoid that. The async backends `moq-tokio` wraps (qmux, iroh, noq) have -no credit query in `web-transport-trait`; prefer adding one upstream in +no credit query in `web-transport-trait`, which `moq-tokio` needs to +implement that method for them; prefer adding one upstream in moq-dev/web-transport over leaving them waiting, and say in the PR which backends fail fast. Propose the error variant's name in the PR. Public API: one new `Error` -variant, and possibly a transport-trait method. Wire: none. +variant, and possibly a `transport::poll` session method. Wire: none.