Meeting 2019-05-01

RT:

  • mayhemer: https://bugzilla.mozilla.org/show_bug.cgi?id=1547556; should we implement this?
  • mayhemer: https://bugzilla.mozilla.org/show_bug.cgi?id=1544102; anyone can think of what could cause this?
  • mayhemer: https://bugzilla.mozilla.org/show_bug.cgi?id=1540731; pretty awesome changes to how IPC actors are refcounted, behold! or see my conversation with nika

mayhemer

BT: talked to JS people to use it to examine off-main-thread parsing effect on page load, result: more work on BT is needed, but it is definitely worth

BT: actively looking for help to instrument more layout code

BT: more ideas on analytical approach to BT data

logan: added full support for zip loading, even when the zip is hosted on web (allow-origin:* needed!)

Bug 1536511 - Implement stale-while-revalidate c-c response directive handling in HTTP cache validation code; finally landed and sticks

Bug 1542322 - Stale-while-revalidate background requests is made even after closing/canceling the load of the page that requested the resource: decide on priority, this release or later? IMHO, not feasible for this release cycle, but also not easily upliftable

Bug 1546780 - Investigate necko priorities for speculative loads: watching this one closely!

reviews, discussions

Became peer of two modules: cookies and permissions

Was on triage: we have unassigned P1s:

1540759

1528481

1547266

Michal

  • 1546975 - Cache SSL resumption tokens in necko
  • 1548472 - Collect telemetry about how is the cache split by the content type
  • 1544655 - WPT on Geckoview frequently crashes in some websockets tests
  • 1547797 - Entry size limits should be different for different content-types

Andy

  • Neqo (QUIC)
    • More flow control message work - don't queue duplicate flow control frames for sending. (Different frames have different scopes -- connection, stream, or stream type -- to be unique within)
  • DoH
    • Mitigating parental controls: nobody likes resolving explicit origins. Selena found an alternative that involves resolving google, which gets redirected to safesearch google. Going to try this myself, fingers crossed.
    • Mitigating split horizon: turning off TRR for non-eTLDs to start. Sounds like this needs per-channel TRR-disable capability --> Necko changes? (new load flag?) Need to meet with Valentin + others to figure out next steps, try to sched. for tomorrow.
    • Looking to do another study to get an idea of just how often we will be turning off TRR if we adopt these or other heuristics. Hopefully we can get an answer quickly. Will be trying to get this underway today.

Valentin

PTO

Bug 1525640 - Let consumers of nsIHttpChannel know if resolving the name was done via TRR (relanded)

resumimg work on:

Bug 1536744 - Make NS_NewURI work off main thread and remove nsIProtocolHandler.newURI

Bug 1527314 - Expand Cross-Origin-Opener-Policy tests

Bug 1533877 - Write WPT tests for Cross-Origin header

Bug 1518730 - TRR: Firefox prefers IPv4 no matter what

Junior

Rebase larch

Bug 1547389 - support HttpTrafficAnalyzer for socket process

Bug 1547025 [socket process] prevent creating Child-to-Socket IPC when socket process is disabled

FATAL /home/roc/rr/rr/src/replay_syscall.cc:1144:handle_opened_files()

Kershaw

Bug 1547705 - [socket process] Race between canceling the channel and send ODA to child process

r?

Bug 1509823 - [socket process] Fix HTTP cache data writing

Cache should be already working

Bug 1497249 - [socket-proc] Connection stickiness handling for upgrade observers

wip

Bug 1545226 - Some gtests in netwerk/test/gtest are not run on Android

Our test harness on Android doesn’t support to load external files yet.

Bug 1542593 - PNeckoChild::SendPUDPSocketConstructor called off main thread

r?

Dragana

neqo

The secure proxy project

The socket process project