Participants

Alex Christiansen, Doug Creager, Benjamin De Kosnik, Nic Jansma, Thomas Kelly, Will Hawkins, Nicolás Peña, Yoav Weiss, Ilya Grigorik, Steven Bougon

Administrative

Next call on July 10th, 10AM PST — design call, we’ll move the invite.

Agenda

Philippe: working on new dashboard for the group

… we’ll review the new dashboard on next call

… still under construction

Respec failures

Yoav: we have a number of ref errors in our specs

… this is a recurring problem where we have failures in published specs

… any action or things we should consider?

Philippe: the problem is that Marcos is making changes and ground is moving from under us

Ilya: can we snapshot or add a build step? That will enable us to revalidate the output before its in production

Philippe: Even if we freeze the respec version, there are still server side services that can change

Yoav: Yeah, this time around it was Shephard failures. But with Respec it causes production failures. Freezing the HTML output sounds like a low overhead way of resolving this problem

Nicolás: so we want to freeze what we serve instead of current flakiness

Ilya: It’s not flakiness, it’s improvements. But changing the flow will enable us to control it
AI: investigate adding a build step for respec, we have some existing examples

Performance Timeline

How does "queue a PerformanceEntry" algorithm interact with other specs?

Nicolás: seems like missing example, we can share?

… maybe a question about a flag that was removed? If so, should be non-issue.

Will: Josh is in Toronto office, can clarify with him

AI (Nicolás, Will): we need to clarify what the question is.

"Relevant performance entry tuple" usage is still broken

Nicolás: this is a simple fix, just haven’t got to it.

… my name is on it, I’ll get to it soon

WPTs

Yoav: Chrome and Edge are green

… Firefox has a few failures, looks like supportedEntryTypes failures

Nicolás: there’s a few User Timing L3 dependencies that are failing, which is unfortunate

Will: I will look at the entry type failure

Ilya: re, UT failing, is there a way for us to avoid this?

Nicolás: we don’t have another way to create entries in PerformanceTiming

… we always have a way to have a dependency

Ilya: FF supports UT, why are we failing?

… the harness test checks for L3, which fails

… is there a way to test against IDL version of LN version?

Philippe: no

Ilya: no resolution, these failures are not failures

… what about Safari failures?

… there is no PerformanceObserver support

… but there are some getEntriesByType failures

Alex: we’re aware of the issue, but won’t get to it for another month or so

Ilya: great, wanted to make sure you’re aware of it

Page Visibility

Update respec version as EventHandler currently has a respec error

Yoav: still some failures

Philippe: if you see ref errors, please assign those to me

PR: Allow a user agent to return hidden when obscured

PR: Correct reference of Page Visibility to top level browsing context

PR: Add external hooks to now {visible,hidden} algorithms

Yoav: we need reviewers

AI: Yoav, Alex, Will to review

WPTs: All green other than prerender

Philippe: we already removed prerender in the publiced version

… at the f2f we also removed it from the editor draft

Yoav: ok, so we need to remove the test — correct?

Philippe: yes

AI: Yoav will open issue to remove the WPT test

Ilya: there are WPT failures for Safari

Yoav: they’re timeouts, unclear what the failures are?

Alex: looking into that

requestIdleCallback

PR: Clarify timeRemaining() algorithm.

Yoav: I need to do a second sweep, should be close to landing.

Deadline definition

Yoav: discussed at the f2f with Ryosuke and think we have general agreement

… don’t have proposed wording, but think we have the general concepts

Ilya: what are the next steps?

Yoav: there is a PR 78 in flight, I will update to capture discussion in 71

WPTs

Green other than Safari, which does not (yet) implement rIC.

NavigationTiming

Yoav: a couple of issues that “just need work”, missing tests

Ilya: we need to rename 99 to capture same origin

Yoav: yes, will do

WPTs - almost green other than unload/redirect changes

Ilya: we recently updated TAO to same-origin

Yoav: we’re green in Chrome

Will: on my radar, this is next on my list

Nicolás: Edge is not green?

Yoav: change probably hasn’t landed yet, we’ll have to check

Alex: don’t have a bug on it

Yoav: there’s a bunch of failures, but I wonder if PerformanceObserver is the issue that’s causing the failures?

Ilya: could safari add PerformanceObserver? :)

Alex: it’s a possibility