2020-05-14 meeting notes

[valenting]

Round table:

• what should we do with 404 response with no content ?
• 1604056 kernel panic on LineageOS
• HTTP/2 issue when toggling VPN state

Next on triage duty: Andy

[valenting]

Triage / Outreachy

[Landed]
Bug 1572513 - Abort and BFCache handling for COEP headers
Bug 1614988 - COOP+COEP data URL tests timing out
Bug 1624090 - We should make possible to turn on neqo logging in about:networking
Bug 1635899 - We should cancel TRR::mTimeout in OnStopRequest.
Bug 1634462 - Check doh-rollout.mode in TRRService
Bug 1635566 - TRR: Perform a case-insensitive match for the host name

[Working on]
Bug 1636888 - Allow specifying entire rust modules in MOZ_LOG
#528 - Implement all qlog probes
Bug 1595848 - Intermittent COOP/COEP tests

[agrover]

Neqo (7, 32, 52)(-2,+1,+1):

• Project Plan 0.2
  • Q: What are conditions to re-enable in Nightly? A: Serious quic-network-sim (QNS) progress and no prio=high issues remaining
• #101 Can't xfer data with winquic - Reproduced with another QUIC impl. It's not us, it's them. (It's only HTTP 0.9 so not critical tbh.) Closed.
• Google interop: They report an issue when initial CID != 8 bytes. We do this. They estimate 3 weeks to fix & roll out. No temporary workaround on our side planned.
• Worked on neqo_glue changes for header being returned in an event.
• Reviewed many PRs from mt, dragana, & vonasek. Wow!

Next:

• Even more PRs to review!
• Work on glue changes for next Neqo->Gecko merge
• Maybe Persistent congestion interval start time

[vonasek]

• 1630000 Hitting Back sometimes shows old cached content
• neqo-client: Handle UdpSocket::send() errors mozilla/neqo#619
• assertion failed: earliest > now mozilla/neqo#621
• make image smaller neqo-qns#2
• Quic network-simulator
• 1635935 HTTP/2 issue when toggling VPN state
• 1637947 network ID: link status isn't updated in NetlinkService::OnAddrMessage()

[mayhemer]

• preload
  • landed: M1 and half of M2, still disabled
  • dogfood at will!
  • missing: fetch(), XHR, Link:
  • on track: perf and integrity testing
• landed: MOZ_LOG=events:1 and logan
• planned: MOZ_LOG=events to add IPC
• updated: our triage helper
• ni?, r?, here and there

[KershawChang]

Landed:

• Bug 1549323 - [socket-process] Check whether SSL_ClearSessionCache() is called on the socket process.
• Bug 1636999 - [socket process] handle the case that connection manager is used when socket process is not ready
• Bug 1636930 - [socket process] Make sure http response head can be serialized and deserialized properly

Ongoing:

• Bug 1539819 - [socket process] TRR is not working in socket process
• Bug 1637648 - [socket process] Forward observer notifications in nsHttpHandler to socket process
• Bug 1616014 - DNSListenerProxy::mListener should not be a main thread pointer holder

[ddragana]

Finishing HTTPSSVC design

Done

• PR 601 Close a connection if critical streams are reset or a stop_sending is received
• PR 598 - Improve table.rs
• PR 611 - Add an encoder's max_table_size.
• PR 612 - Remove total_num_of_inserts, use table.base() instead
• PR 614 - Refactor qpack errors
• PR 617 - Add a reference to a qpack dynamic table entry as soon as it is referred in a header block encoding
• PR 616 - The qpack encoder uses stream_send_atomic to send instructions

Next

• Issue 459 - qpack: Fix Header Block Prefix
• Issue 300 - Implement server push, client side
• Issue 417 - telemetry

[JuniorHsu]

• [land] Bug 1632045 - Content-Type and related headers not stripped for certain redirects when using XMLHttpRequest
• Bug 1636572 - [webvtt] The testcase fails if we allow sending onStartRequest via pHttpBackgroundChannel
• Bug 1633935 - Allow sending onStartRequest via pHttpBackgroundChannel
  • P2 let HttpChannelBackgroundChild know the sink events before OnStartRequest