W3C home > Mailing lists > Public > public-web-perf@w3.org > May 2018

Web Performance Design notes 2018-05-17

From: Philippe Le Hégaret <plh@w3.org>
Date: Thu, 17 May 2018 15:48:54 -0400
To: "public-web-perf@w3.org" <public-web-perf@w3.org>
Message-ID: <a50c7546-4713-2689-f533-ec9915c9c6f3@w3.org>
               Web Performance Working Group Teleconference

17 May 2018


           yoav, plh, Tim, Nic, nolanlawson, philipwalton, garrett





      * [2]Topics
      * [3]Summary of Action Items
      * [4]Summary of Resolutions

    plh wants to add NavTiming2 to the agenda

Topic: F2f?

    Nic: F2F at Fluent? Was good in past years

    Todd: Not planning to be at Fluent. perhaps try a virtual F2F
    as a way to make forward progress

    NicJansma: Would work for me. Could be hard but we can give it
    a try

    yoav: Having Apple and Mozilla in the room is pretty

    yoav: Including all primary implementors makes these
    meetings much more valuable.

    AI: Todd to schedule a virtual full day call about a month from
    ... Yoav to tag external folks that are critical for some of
    the RT issues and schedule time within that day to discuss
    those issues

    Todd: Charter as written is appropriate. Renewing the charter
    makes sense. The chairs will coordinate next week

Topic: User Timing

    philipwalton: User Timing - 1 outstanding issue




       [6] https://github.com/w3c/user-timing/issues/29

    <toddreifsteck> User Timing Issue

       [7] https://github.com/w3c/user-timing/issues/29

    yoav: seems similar to the issue cvazac ran into when writing
    Server Timing tests

    cvazac: Does look similar

    toddreifsteck: we can create a stand alone test instead of
    using the non-existing infra

    assigned to npm

    plh: are we passing the tests for PerfObservers on user timing?


       [8] https://wpt.fyi/user-timing

    plh: the tests should be in the user-timing directory. I'll
    check where they actually are


       [9] https://wpt.fyi/performance-timeline

    po-mark-measure is green in 3 implementations

    toddreifsteck: are we taking user timing to CR?

    Resolved: User timing is moving to CR

Topic: HR Time

    <toddreifsteck> [10]https://github.com/w3c/hr-time/issues/56

      [10] https://github.com/w3c/hr-time/issues/56

    plh: the spec is imprecise on purpose

    toddreifsteck: maybe we need a bit more specific language?

    plh: you tell me

    toddreifsteck: I don't think a more precise language is needed.
    I think implementers can implement

    <plh> [11]https://w3c.github.io/hr-time/#clock-resolution

      [11] https://w3c.github.io/hr-time/#clock-resolution


      [12] https://w3c.github.io/hr-time/#sec-domhighrestimestamp

    plh: the note is no longer accurate. We can remove it, close
    the issue and move to PR?

    toddreifsteck: fair

    RESOLUTION: hr-time-2 to proposed recommendation

    toddreifsteck: and that will update the existing proposed rec

    plh: yeah, we should send a note to the privacy folks and see
    their reaction

Topic: Beacon

    <toddreifsteck> [13]https://github.com/w3c/beacon/issues/58

      [13] https://github.com/w3c/beacon/issues/58

    toddreifsteck: assigning to nolanlawson


      [14] https://github.com/w3c/beacon/issues/59

    toddreifsteck: Brandon will fix this in sometimes in the next
    ... otherwise, I think we're all green
    ... have we sent beacon to rec yet?

    plh: once the tests are green, REC it is

    <xiaoqian> [15]https://wpt.fyi/beacon

      [15] https://wpt.fyi/beacon

    No implementation is fully green

    toddreifsteck: need to ask someone to run it on all the latest
    dev versions, after the tests are fixed
    ... AI send an email to report on test results on dev browsers
    after tests are fixed

Topic: Resource Timing

    NicJansma: We used to have a couple of branches. Decided to not
    have a split between the v2 and gh-pages

    now gh-pages is v2 and v2 branch is gone

    At the point we actually ship v2, we will re-branch

    2 open PRs


      [16] https://github.com/w3c/resource-timing/pull/155

    Proposal to change the minimum buffer size to 250

    not much opposition

    not a requirement, as browsers can have a lower size

    but it will help many analytics cases

    toddreifsteck replied that it look good for Edge

    digitarald will check with Moz team

    <tdresser> tdresser taking over as scribe


      [17] https://github.com/w3c/resource-timing/issues/110

    <tdresser> 16 open L2 issues in resource timing


      [18] https://github.com/w3c/resource-timing/issues/148

    yoav: 110: It felt like we should put a note in the
    fetch spec, but in practice looking at the fetch spec, it
    didn't make sense, so we just closed the issue.

    yoav: 148: Merged PR and closed issue, getting rid
    of mention of default actions.


      [19] https://github.com/w3c/resource-timing/issues/70

    yoav: No one implements the spec here.

    yoav: No security threat here, service workers can
    get the same information.

    yoav: The spec says we should exclude resources
    fetched by cross-origin stylesheets fetched with no-cors

    yoav: No one does this, we should update the spec to
    reflect this.

    <tdresser> todd: Do TAO headers matter here?

    yoav: It's not a question of TAO, it's a question of
    whether the URL is shown in the Resource Timing entry.

    <tdresser> todd: The conclusion that we should expose these
    seems reasonable to me.

    <tdresser> digitarald: I'll check with other moz folks.


      [20] https://github.com/w3c/resource-timing/issues/122

    <tdresser> yaov: We don't have an ALPN for http1.0.

    yoav: I'll add a note explicitly mentioning http/0.9
    and http/1.0.


      [21] https://w3c-test.org/navigation-timing/po-navigation.html

Topic: Navigation Timing

    plh: NavTiming2 is not implemented by anyone but chrome. Edge
    doesn't count because no PerfObs

    toddreifsteck: so you choose to tie this spec to perf timeline

    plh: the tests don't pass, so...

    toddreifsteck: you're tying 2 specs that are not joined
    ... on Edge 17 we will match chrome on all properties other
    than the bytes properties (probably at Edge 18)

    this time next year we'll ship exactly what chrome is shipping

    Firefox is very close to chrome

    plh: that spec is not being implemented rapidly

    toddreifsteck: Chrome implemented the breaking change, now
    other UAs are slowly following

    that's just how breaking changes get deployed


      [22] https://wpt.fyi/navigation-timing/nav2_test_redirect_server.html

    panagiotis: will look at the test and see how far Firefox are

    toddreifsteck: Chrome shipped almost a year ago, right?

    plh: yeah

    toddreifsteck: afaik, no issues reported
    ... did the change of name from document to url break anything?

    tdresser: will check with Ilya

    toddreifsteck: so right now, no 2 implementations. shortest
    path is Firefox fixing a bug
    ... do we know where safari is?

    yoav: both tests don't seem to pass on STP

    toddreifsteck: next call on 31 may 11am

Summary of Action Items

  Todd to send out request for beacon results for in development 
browsers after tests are fixed.

  Todd to schedule a virtual full day call about a month from now

Summary of Resolutions

     1. user timing 2 to candidate recommendation
     1. HR time 2 to proposed recommendation

    [End of minutes]
Received on Thursday, 17 May 2018 19:49:17 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 17 May 2018 19:49:17 UTC