[minutes] 20150909 Web Performance

Available at:
   http://www.w3.org/2015/09/09-webperf-minutes.html

Text version:


Present
     plh, yoav, ilya, todd, brandon
Regrets
Chair
     Ilya
Scribe
     plh

Contents

     Topics
         requestAnimationFrame
         Obsoleted EDs
         requestIdleCallback
         Resource Timing
         Navigation Timing
         Page Visibility
     Summary of Action Items

requestAnimationFrame

Todd: didn't review the HTML spec to see if they have everything there 
but not a blocker

Resolved: Published a Working Group Notes pointing to the HTML spec for RAF
Obsoleted EDs

plh: took care of that a week ago. should be fine now
... (if not, scream)
requestIdleCallback

<igrigorik> https://github.com/w3c/requestidlecallback/pull/16

plh: request for FPWD sent

Ilya: where should the test PR go?

plh: indeed, should go to WPT

<igrigorik> https://github.com/w3c/resource-timing/issues/17
Resource Timing

Todd: I'll look into RT/#17

<igrigorik> https://github.com/w3c/resource-timing/issues/12

Todd: no pushback from our security folks

Ilya: meaning exposing the timestamp isn't an issue?

Todd: with NEL, would we want this to be exposed?

Ilya: Resource fetches should be reported to the page
... while error logging sits on top of that, ie ifr the fetch fails, 
what's to know about it

https://github.com/w3c/resource-timing/issues/21

Ilya: the web platform has a lot of transparent fetches (redirects, etc.)
... but would be good to surface
... today, you can infer than a redirect happened
... if we can surface the number of redirects, is that a bad thing?
... for perf, we don't have visbility into this
... despite the impact
... we need to clarify fetch first before we can define the behavior

Yoav: the preflight case would be safe to expose
... while redirects are more risky (don't have a specific case in mind)

Ilya: we want to simplify and improve our reporting in RT
... if we can bundle them, it would make our model much easier

Todd: the key in security issue is exposing data from the URL
... if the data wasn't exposed previously, that would be the risk

Ilya: we could choose not to reveal the URL, only the fact that a fetch 
happened
... at least, you'll get your waterfall correct

Todd: we probably need to kick the fetch thread

Ilya: I'll follow
Navigation Timing

Todd: I was looking into this and didn't find the case where we would 
have multiple documents

https://github.com/w3c/navigation-timing/issues/1

<ToddReifsteck> TAO for Nav timing-- 
https://github.com/w3c/navigation-timing/issues/20

(summary: assign the issues to plh and get him to do the 
navigation-timing rewrite :)

<igrigorik> https://github.com/w3c/page-visibility/issues/8

http://w3c.github.io/page-visibility/#hidden-attribute

<yoav> Here's the relevant Blink work: 
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/_SRHebxivJs

http://www.w3.org/2014/10/pv2/iframe-hidden.html
Page Visibility

https://github.com/w3c/page-visibility/issues/12

Todd: I'll create an issue to see if we can remove hidden

http://microformats.org/wiki/rel-prerender

Ilya: we define the relationship in the wiki
... we'll put more details in resource hints
... but we never talk about prerender mode
... we never define it

plh: I'll open an issue on resource hints to define prerender mode

Ilya: I'll take care of it

[adjourned]

Received on Monday, 14 September 2015 14:46:33 UTC