[minutes] 20110119 Web Performance WG

Web Performance Working Group Meeting
19 Jan 2011

See also: IRC log<http://www.w3.org/2011/01/19-webperf-irc>

Minutes
http://www.w3.org/2011/01/19-webperf-minutes.html
Attendees
Present
+1.650.214.aaaa, +1.650.704.aabb, Plh, [Microsoft], +1.650.450.aacc, +1.650.253.aadd, +1.650.704.aaee, TonyG, Zhiheng, JasonSobel, JamesSimonsen, NicJansma, KarenAnderson, ArvindJain
Regrets
                JasonWeber
Chair
ArvindJain, JasonWeber
Scribe
AndersonQuach
Contents

  *   Topics<http://www.w3.org/2011/01/19-webperf-minutes.html#agenda>
     *   Review and approve Navigation Timing conformance tests.<http://www.w3.org/2011/01/19-webperf-minutes.html#item01>
     *   User timings, proposed updates and changes.<http://www.w3.org/2011/01/19-webperf-minutes.html#item02>
     *   Summary<http://www.w3.org/2011/01/19-webperf-minutes.html#item03>

         i.            We'll update the tests according to what was discussed in this email thread and move them to the approved folder.

a.        Anderson will rename of mstestharness.js to webperftestharness.js then coordinate with the owners of testharness.js to incorporate the changes to the central harness.

b.       Propose the addition of a new test that tests the existence of optional attributes

        ii.            Zhiheng will update the User Timing specification as discussed today.

      iii.            Anderson, Nic and Karen will look into novel functional requirements that will help web developers determine time to when an element is displayed on the screen and the time when an element can be interacted with to help seed further design discussions on User Timings.

      iv.            All, continue review and gathering feedback on Resource Timings and User Timings.

________________________________
Review and approve Navigation Timing conformance tests.

AndersonQuach: We should encourage non-members to submit tests. Where non-members post it somewhere and a WG member takes point to push it through the process.

plh: Yes, other groups do this.

<plh> public-html-testsuite

<plh> http://lists.w3.org/Archives/Public/public-html-testsuite/

plh: we have two layers in the harness, we have testing frameworks and a javascript layer.
... test harness is becoming more and more stable, the framework is not as stable.

NicJansma: Let's start to get these tests approved and the test harness.

AndersonQuach: Let's go with a two phase approach, get the tests approved and check them into the approved folder and then work with the public-html-testsuite to get the right changes into the central testharness.js

KarenAnderson: what's the timeframe like to get changes in place.

plh: javascript changes turn around is around two weeks. the framework may take more.

<plh> http://test.w3.org/webperf/tests/submission/Microsoft/NavigationTiming/test_navigation_attributes_exist.htm

<plh> approved

<plh> http://test.w3.org/webperf/tests/submission/Microsoft/NavigationTiming/test_navigation_redirectCount_none.htm

<plh> approved

<plh> http://test.w3.org/webperf/tests/submission/Microsoft/NavigationTiming/test_navigation_type_enums.htm

<plh> http://test.w3.org/webperf/tests/submission/Microsoft/NavigationTiming/test_timing_attributes_exist.htm

NicJansma: what's should be the approach for a value that does not have to exist? but conforming to the definition.
... It does exist, nop, if it does exist it passes.

TonyG: I'm assuming we don't add anything with the existence.

AndersonQuach: clearly mark the optional tests are non-normative.

plh: optional tests do not count towards the normative pass/fail.

NicJansma: is there a skip result in the test harness for optional.

AndersonQuach: next steps are, add an optional secureConnection test and in the processing model we'll have a conditional test for the secureConnection model.

All: all four tests approved.

move to agenda 2
User timings, proposed updates and changes.

Zhiheng: I'm half way updated the user timing spec. i had some questions. seems to me, all these functions in user timing can be easily implemented by a framework. we are not adding anything new.
... question is it better as a framework or in the DOM?

AndersonQuach: the main advantage is the high-precession than most javascript engines.

NicJansma: The two other benefits, it's one less library the website to include in the js files, what ever we push into the browser we can be more efficient with it.
... faster searches, and also the standard marks, other frameworks can key off it and analytic engines can utilize this.

AndersonQuach: The motivation is when an element first painted to the screen and the time when an element is interactive.

TonyG: we've heard developers want metrics to know when an ad displayed on the screen.

NicJansma: We would need to investigate the feasibility, but this is what web developers have been seeking.

TonyG: would need to make it distinct, above the fold and below the fold.

JasonSobel: do we know how far apart is onload from painting. my hypothesis painting is a small portion of the end to end time for an element. i want to know when this shows up on the screen, are resource timing and navigation timing 90-99% of the time.

TonyG: that might be true, for a trival page with a small fixed delay. the use case, some complex page with ads and with navigation links and main content textual article. i want to know when did my article show up on my screen. that's blocked by stylesheets and script.

NicJansma: It depends the additional work the browser has to do before it paints.

Zhiheng: allow developer and browser to provide useful performance metrics.

set logs world-visible
Summary

i. We'll update the tests according to what was discussed in this email thread and move them to the approved folder.

a. Anderson will rename of mstestharness.js to webperftestharness.js then coordinate with the owners of testharness.js to incorporate the changes to the central harness.

b. Propose the addition of a new test that tests the existence of optional attributes

ii. Zhiheng will update the User Timing specification as discussed today.

iii. Anderson, Nic and Karen will look into novel functional requirements that will help web developers determine time to when an element is displayed on the screen and the time when an element can be interacted with to help seed further design discussions on User Timings.

iv. All, continue review and gathering feedback on Resource Timings and User Timings.