Meeting 2020-05-14

Round table:

Next on triage duty: Andy

Valentin

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

Andy

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:

Michal

  • 1630000 Hitting Back sometimes shows old cached content
  • https://github.com/mozilla/neqo/pull/619
  • https://github.com/mozilla/neqo/issues/621
  • https://github.com/mozilla-necko/neqo-qns/pull/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

Kershaw

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

Dragana

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

Junior

  • [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