<scribe> scribe: igrigorik
todd: one issue remaining on HR-time
dale: PR is in place, should be good
... tests are in pretty good shape, but would like to get more eyes on it
... fails in FF today, potentially due to implementation bug
... added some details in the test PR
... safari is not testable today because it's a lot more aggressive about how popups are handled
... don't if/how to proceed there
spec update https://github.com/w3c/hr-time/pull/54
tests https://github.com/w3c/web-platform-tests/pull/8488
RESOLUTION: spec text is good, merged.
[tests: waiting for Boris to comment on Dale's repro case]
todd: we have two passing implementations, good to publish ; let's wait for Boris to followup.
Nick: auditing v3 branch vs v2 differences
... also going through open v2 issues
Issue #120 https://github.com/w3c/resource-timing/issues/120
scribe: discussed at TPAC, one of the followups was to gather some RUM data
... we gathered bulk data for couple customers
... we did find some instances of timestamps in non-standard order
... most of those had requestStart < connectEnd
... plus some weird data, e.g. requestStart having a value but others being zero
... not sure if its browser bugs or bugs in our collection
... near 70% of weird ordering is for QUIC
todd: Edge has attempted to ship TFO in multiple release, so far we've had to back off each time
... don't have enough backoff to detect edge cases
... still WIP and aiming to make it live
... could next step here be to summarize per UA and see if we can isolate some repro cases?
Nic: my suggestion is that some timestamps should have hard ordering, but in others we may want to allow some flexibility
<scribe> ACTION: Nic to followup on issue with next step
Issue #130 https://github.com/w3c/resource-timing/issues/130
Nic: looking into how RT works with media use cases
... documented in the issue differences between browsers today
... e.g. video can trigger poster image
... how should it be reported? spec suggests "video", which makes sense.. but it's not obvious that image is being downloaded.
... also, what about <source>: today it shows up as other, video or blank
todd: let's label each of these use cases and clarify the behavior in current L2 spec
... for this case, it sounds like we need to file some implementation bugs and clarify the spec
Nic: I'll followup on the issue.
Issue #122 https://github.com/w3c/resource-timing/issues/122
Todd: we can add a normative MAY for 0.9; we don't need to fight the battle to register it with IETF
Tim: looked into usage of UT timestamps in the wild
... found that NT1 timestamps are used enough that we can't deprecate them outright
... we'll have to support that in L2
Philip: for L3, we didn't see any usage of objects so we should be in good shape there
... in NT2, using NT1 timestamps has different time origin
Todd: right, we need to map these to NT2 concepts
... in the spec, and define that clearly in the spec
Philip: what about Worker case?
ig: in worker we don't expose NT
Todd: we should probably just throw, NT in worker doesn't make sense
Issue #79 https://github.com/w3c/navigation-timing/issues/79
Todd: Andy is calling out that beforeUnload execution
Ilya: I don't believe we can expose this due to privacy
AI: Ilya to followup on the thread, suggesting WONTFIX
RESOLUTION: next call will be on Jan 5th