Web Performance Working Group Teleconference

Page Visibility

<ToddReifsteck> https://github.com/w3c/page-visibility/issues/18

Todd: current state is that Firefox implements the normative state

Ilya: I missed the point on state change on unload
... everyone except FF missed it

Todd: so should the spec stay the same or not?

Ilya: we have a mixed of implementations now
... I see Boris' point that subscribing to 2 events isn't convenient
... we could say that everyone should implement the spec
... I didn't find any reason in chrome on why it wasn't implemented

Yoav: guessing that safari is using the same old implementation from webkit
... so same reason

todd: I'm tracking this down in MS and my guess it's an omission
... will keep looking

Ilya: we don't see any issue in implementing but it's low priority

todd: agreed. it's silly to cut it when we all shipped the spec together

Ilya: to clarify: this is about firing the hidden transition on unload. we would still remove the unload state
... the text is there but I can make it explicit

Todd: sounds good to me
... ditto it will be low priority in edge as well
... 6 to 9 months to get to it

Ilya: I'll add an additional note to the spec

plh: should we wait to publish a new rec or wait for implementations to catch up?

<ToddReifsteck> https://github.com/w3c/requestidlecallback/issues/31

requestIdlCallback

<igrigorik> https://github.com/w3c/requestidlecallback/pull/32

Ilya: Ross addressed most of the feedback in his PRs

<igrigorik> https://github.com/w3c/requestidlecallback/pull/35

Ilya: lots of clarification but no material change to the API
... the changes look good to me

Todd: I need to make sure the processing model got fixed, ie not tied to Blink model

Ilya: I believe we did but open a separate issue if not

Todd: ok, I'll look at the PR and see to merge

Resource Timing

Todd: almost ready to send some tests to plh
... and we'll go from there

Primer

<ToddReifsteck> a. When will next Primer draft occur? ANSWER: Primer repo-- https://github.com/w3c/perf-timing-primer Primer can be viewed at-- http://w3c.github.io/perf-timing-primer/index.html b. What are next steps for linking/feedback?

plh: next step is for people to look at it and see if we can publish as a working group note

todd: where will it be linked from?

plh: wherever we want

todd: ok, we'll come back to it in 2 weeks

HR Time

<ToddReifsteck> a. Can we merge?--Current Document https://github.com/w3c/hr-time/pull/14

plh: I think we should merge.

Todd: ok

ilya: sound good

plh: ok, I'll squash the edits and then do a merge

Preload

https://github.com/w3c/preload/issues/36

<ToddReifsteck> a. Allowing empty/invalid ‘as’-- https://github.com/w3c/preload/issues/36

Yoav: 2 issues in it
... should we allow invalid or empty as values?
... we could special case empty values but seems weird
... main concern with invalid values is that devs will set the wrong values and will have a hard time figuring why their resource is low priority
... we can mitigate that on the console

Ilya: one of the criteria is to allow to do fetch with passing as. we're tied to it closely.
... the default is to do a low priority fetch
... so we could reject invalid types, but would we fail the fetch?
... the UA could issue a warning but shouldn't reject it
... and we should allow empty value, as a declarative XHR

Yoav: ok. now for CSP directives

Ilya: in previous iteration, we have many more values for as
... CSP is not to block request but to block consuming responses
... not ideal but we don't break anything like that
... if we want to enforce CSP on preload we have some choices

yoav: besides the impl issue, the problem with the generic fetch, devs will have to declare their resources twice, ie declare them up-front
... but if we want the various types, we could have scriopt, worker-script, serivecxe-worker, so we can set the context

ilya: preload itself is not subject to CSP. there is no csp policy that covers it
... I believe Mike West is ok with that

yoav: but if you want to prevent leaking data
... if you exempt preload from that, you cannot prevent data leaks
... it could weaken CSP

ilya: ok, we should keep iterating
... maybe we need a different mechanism to enfore csp on preload

yoav: I think we could extend type
... I'll open a new issue on github

HR Time

<ToddReifsteck> b. Need a changelist without translateTime so we can publish REC https://github.com/w3c/hr-time/issues/16

plh: I'll make a branch
... one is a CR with translateTime and then, once moving to PR, I'll publish a HR Time 3

Performance Timeline

<ToddReifsteck> a. Added performance entry buffer-- https://github.com/w3c/performance-timeline/pull/49

Todd: I'll need to review this

<ToddReifsteck> b. Any updated thoughts on buffering for PerformanceObservers? http://www.w3.org/2015/10/webperf-tpac2015-minutes#ptpo-conclusion (Should we move this to a GitHub issue?)

Plh: I'll create a github out of our TPAC discussion

next meeting

Todd, plh, and yoav will be available on 11/25

Todd will check with Ilya [and indeed, Ilya will be able to make it]