W3C home > Mailing lists > Public > public-web-perf@w3.org > October 2015

[minutes] Web Performance 20151021

From: Philippe Le Hegaret <plh@w3.org>
Date: Wed, 21 Oct 2015 16:35:07 -0400
To: public-web-perf <public-web-perf@w3.org>
Message-ID: <5627F6FB.7030005@w3.org>
Available at
  http://www.w3.org/2015/10/21-webperf-minutes.html

Text version:

                             Web Performance

21 Oct 2015

    [2]Agenda

       [2] 
https://lists.w3.org/Archives/Public/public-web-perf/2015Oct/0007.html

    See also: [3]IRC log

       [3] http://www.w3.org/2015/10/21-webperf-irc

Attendees

    Present
           Yoav, Ilya, plh

    Regrets
           Todd

    Chair
           Ilya

    Scribe
           plh

Contents

      * [4]Topics
          1. [5]Clarify {transfer,encodedBody,decodedBody}size
          2. [6]Page Visibility
          3. [7]"Red warning" for TR/hr-time
          4. [8]Drop prerenderSwitch from nav timing 2
          5. [9]clearMarks
          6. [10]TPAC
      * [11]Summary of Action Items
      __________________________________________________________

Clarify {transfer,encodedBody,decodedBody}size

    - [12]https://github.com/w3c/resource-timing/pull/41

      [12] https://github.com/w3c/resource-timing/pull/41

    Ilya: based on implementation feedback: clarifying what those
    values should report when you don't hit the network
    ... need to address the feedback from Yoav

    <igrigorik> Pull:
    [13]https://github.com/w3c/resource-timing/pull/41

      [13] https://github.com/w3c/resource-timing/pull/41

    Ilya: we can reuse the concepts from fetch
    ... I'll make it clearer
    ... distinguishing between not allowed
    ... if we don't pass the timing check algo, we report 0

    Yoav: it might be a breaking change
    ... what if the resource was actually 0?
    ... the distinction with not allowed could be meaningful

    Ilya: that's the behavior we have for other attributes (we
    report time 0)
    ... if you hit the network, you'll never get non-0 for
    transferSize
    ... you can only know the transferSize if we pass the TAO
    header
    ... if we don't, we always report 0
    ... so if we report not allowed, then we may be saying too much

    Yoav: I wonder what this would expose... TAO aren't user
    specific
    ... the privacy concern would be if the server sends TAO
    headers depending on the user being logged in or not
    ... not sure if TAO is used that way

    Ilya: for consistency, if TAO disallows timing values, we
    report 0

    Yoav: I agree but we should try to change this

    Ilya: if we make nullable or undefined, we could break pages
    ... we should look into this in a different issue

    Yoav: I'll open one

    Ilya: I'll keep it as 0 for now

    <scribe> ACTION: Yoav to open an issue on RT regarding TAO
    information and attributes [recorded in
    [14]http://www.w3.org/2015/10/21-webperf-minutes.html#action01]

    <trackbot> Created ACTION-161 - Open an issue on rt regarding
    tao information and attributes [on Yoav Weiss - due
    2015-10-28].

    Ilya: with the current, if any of the requests failed during
    redirects, we set the attributes to 0.

    Yoav: if the conditions don't match and we're not adding the
    http overhead.... I guess that's ok.

    <scribe> ACTION: Ilya to update pull 41 with network fetches
    [recorded in
    [15]http://www.w3.org/2015/10/21-webperf-minutes.html#action02]

    <trackbot> Created ACTION-162 - Update pull 41 with network
    fetches [on Ilya Grigorik - due 2015-10-28].

    <igrigorik> FF bug:
    [16]https://bugzilla.mozilla.org/show_bug.cgi?id=1154309

      [16] https://bugzilla.mozilla.org/show_bug.cgi?id=1154309

    Ilya: firefox is implementing right now. hoping Q4 for chrome
    but no one is committed to it.

Page Visibility

    <igrigorik> [17]https://github.com/w3c/page-visibility/pull/16

      [17] https://github.com/w3c/page-visibility/pull/16

    Ilya: I was set to remove unloaded, but I found some issues
    ... we define both visible and the visibilityState with
    different algorithms
    ... so we should fix that
    ... I'm proposing we drop the setting but run the algorithm
    instead

    Yoav: I'm guessing the implementation will set the variable
    instead of running the algo
    ... as long as we're identical in terms of functionality, we
    should be fine

    Ilya: prerender is mandatory instead of optional
    ... everyone implements it
    ... Safari is very aggressive in using prerender
    ... if you paste in the address bar, it will trigger
    prerendering based on predictions

    yoav: as long as clearing the cache will also clear
    predictions, you should be fine

    Ilya: next step is to keep circulating the PR
    ... one outstanding change as well

    plh: user interaction task source?

    Ilya: wasn't sure about that one. will try to get Anne or Boris
    to look at it

    Yoav: we need to separate issue for unloaded
    deprecation/removal

    ILya: ok

    <scribe> ACTION: Ilya to open bug for unloaded in PV [recorded
    in
    [18]http://www.w3.org/2015/10/21-webperf-minutes.html#action03]

    <trackbot> Created ACTION-163 - Open bug for unloaded in pv [on
    Ilya Grigorik - due 2015-10-28].

    <scribe> ACTION: Ilya: to add triggers for changing the value
    of visibility state [recorded in
    [19]http://www.w3.org/2015/10/21-webperf-minutes.html#action04]

    <trackbot> Created ACTION-164 - Add triggers for changing the
    value of visibility state [on Ilya Grigorik - due 2015-10-28].

    <scribe> ACTION: Ilya to check the task source in PV [recorded
    in
    [20]http://www.w3.org/2015/10/21-webperf-minutes.html#action05]

    <trackbot> Created ACTION-165 - Check the task source in pv [on
    Ilya Grigorik - due 2015-10-28].

    Ilya: then we'll have to update resource hints

"Red warning" for TR/hr-time

    Yoav: +1 to renaming idea

    <igrigorik> +1, let's discuss this more at TPAC as well

Drop prerenderSwitch from nav timing 2

    Ilya: nobody implements prerenderSwitch
    ... and it makes it tricky to understand when it is set
    ... so not clear if it is necessary

clearMarks

    Ilya: with performance observers, the buffer logic is expressed
    with simple JS

    <igrigorik>
    [21]https://github.com/w3c/user-timing/pull/13#issuecomment-149
    985067

      [21] https://github.com/w3c/user-timing/pull/13#issuecomment-149985067

TPAC

    <igrigorik>
    [22]https://docs.google.com/document/d/1IPM75UYlPPIPnr5uZMPYwb9
    EcOvquyWTo1uGqHeNafI/edit

      [22] 
https://docs.google.com/document/d/1IPM75UYlPPIPnr5uZMPYwb9EcOvquyWTo1uGqHeNafI/edit

Summary of Action Items

    [NEW] ACTION: Ilya to check the task source in PV [recorded in
    [23]http://www.w3.org/2015/10/21-webperf-minutes.html#action05]
    [NEW] ACTION: Ilya to open bug for unloaded in PV [recorded in
    [24]http://www.w3.org/2015/10/21-webperf-minutes.html#action03]
    [NEW] ACTION: Ilya to update pull 41 with network fetches
    [recorded in
    [25]http://www.w3.org/2015/10/21-webperf-minutes.html#action02]
    [NEW] ACTION: Ilya: to add triggers for changing the value of
    visibility state [recorded in
    [26]http://www.w3.org/2015/10/21-webperf-minutes.html#action04]
    [NEW] ACTION: Yoav to open an issue on RT regarding TAO
    information and attributes [recorded in
    [27]http://www.w3.org/2015/10/21-webperf-minutes.html#action01]

    [End of minutes]
Received on Wednesday, 21 October 2015 20:35:10 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 21 October 2015 20:35:10 UTC