[minutes] 2012-04-11 Web Performance WG Teleconference #68

Meeting Summary:



1.      Navigation Timing

a.       Navigation Timing to PR

We expect WebIDL to go to CR next week. As such, we expect Navigation Timing to move to PR next week.



b.       Navigation Timing Test Status

Per http://www.w3.org/2012/04/navigation_timing_cr_results.html, most UA vendors are failing test_timing_attributes_order.html which makes us question if the test is correct. Karen is investigating if this test needs to be fixed. ACTION-99: Look into test_timing_attributes_order.html to make sure it is correct on Karen. Boris to follow up and make sure Firefox failing two tests is a fault of the test or Firefox.



2.      Resource Timing

a.       Clarifications regarding onresourcetimingbufferfull callback behavior

The spec isn’t very clear on what the behavior should be when the onresourcetimingbufferfull callback isn’t defined. The WG agreed that if this callback isn’t defined, new resources will not be added to the primary resource timing buffer. ACTION-100: Update Resource Timing processing model to take care of the case that onresourcetimingbufferfull isn’t defined on Jatinder.



3.      User Timing

a.       Spec Updates

The following updates to the spec were made by the editor:

                                                               i.      Updates measure method without any parameters to store the duration from navigationStart to the current time, rather than fetchStart to the current time

                                                              ii.      Updated all examples to include sub-millisecond resolution

                                                            iii.      INVALID_ACCESS_ERR exception of measure method typo fixed.



4.      Performance Timeline

a.       Spec Updates

To make the behavior very clear, Section 4.4 startTime attribute definition was updated to return 0.



b.       Navigation Timing attributes in High Resolution Time

There is a discussion on the mailing list if the confusion of including redundant APIs will exceed the benefit of adding the PerformanceTiming attributes in high resolution time in a PerformanceEntry. The WG wants to follow up on this on the mailing list after some more thought: http://lists.w3.org/Archives/Public/public-web-perf/2012Apr/0027.html.



5.      High Resolution Time

a.       Spec Updates

The following updates to the spec were made by the editor:

                                                               i.      Updated DOMHighResTimeStamp definition to take into account durations,

                                                              ii.      Updated DOMHighResTimeStamp definition to recommend microsecond resolution on millisecond units,

                                                            iii.      Added a note saying if sub-millisecond not possible, should return millisecond, resolution in millisecond unit,

                                                            iv.      Added a note saying that time spent not fully active is not excluded from current time.



6.      Page Visibility

a.       Spec Updates

The following updates to the spec were made by the editor:

                                                               i.      Updated Section 4.2 to move non-normative statements into notes and capitalized, MUST and MAY statements,

                                                              ii.      Added prerender as an optional state,

                                                            iii.      Added definition of the preview state,

                                                            iv.      Added an editorial note for Issue 8,

                                                              v.      Added hook to HTML5 unloading processing model in the Page Visibility processing model



b.       Tracking Unload/Suspending Scenarios in ISSUE-8

The specification contains draft text on including the unloading and suspending scenario in the Page Visibility spec, as tracked by ISSUE-8: http://www.w3.org/2010/webperf/track/issues/8. Once this issue has been closed, we can decide to remove or keep the text supporting the unloading/suspending scenarios.


7.      TPAC 2012

The WG believes that though there is value in meeting working group members face to face, the high overhead of a trip to France and unlikeliness of communicating with other working groups at this time, the WG does not want to participate in TPAC 2012 this year. If we need a face to face meeting this year, we can schedule one on the west coast.


Detailed Notes:



Web Perf Teleconference #68 4/11/2012



IRC log: http://www.w3.org/2012/04/11-webperf-irc


Meeting Minutes: http://www.w3.org/2012/04/11-webperf-minutes.html



Attendees

Present for Navigation Timing, Resource Timing and User Timing (4-5PM EST/1-2PM PST)

Jatinder Mann, Philippe Le Hegaret, Alois Reitbauer, Tony Gentilcore, Arvind Jain, Karen Anderson


Present for Page Visibility, Efficient Script Yielding, Display Paint Notifications (4-5PM EST/2-3PM PST)

Meeting cancelled.



Scribe

Jatinder Mann



Contents

Agenda

1.     Review feedback on all specifications

--------------------------------------------------------------------------------
Navigation Timing

plh: The good news is WebIDL is moving to CR, hopefully next week. This means we should plan to move Navigation Timing to PR.

http://www.w3.org/2012/04/navigation_timing_cr_results.html

plh: In these tests, I notice IE9 fails two tests, where as IE10 fails a test that IE9 passes. Comments Jatinder, Karen?

Karen: We discussed the two tests that IE9 failed, and we have fixed that in IE10. We will downport that fix at some point.

<plh> http://w3c-test.org/webperf/tests/approved/navigation-timing/html5/test_timing_attributes_order.html

plh: The test_timing_attributes_order.html is passing in IE10, failing IE9, Chrome and Firefox.

Karen: I wonder if the bug is in the test case, we should look into that.

plh: It looks like there was an action item on updating the test_timing_attributes_order.html that may have not been completed. http://lists.w3.org/Archives/Public/public-web-perf/2012Apr/0020.html

Karen: I will look into it.

Action Karen to look into test_timing_attributes_order.html to make sure it is correct.

<trackbot> Created ACTION-99 - Look into test_timing_attributes_order.html to make sure it is correct. [on Karen Anderson - due 2012-04-18].

<plh> http://w3c-test.org/webperf/tests/approved/navigation-timing/html5/test_timing_attributes_order.html

<plh> http://w3c-test.org/webperf/tests/approved/navigation-timing/html5/test_navigation_type_reload.html

<plh> Fail Reload loadEventEnd(0) == Original loadEventEnd(1334164524631) assert_equals: Reload loadEventEnd(0) == Original loadEventEnd(1334164524631) expected 1334164524631 but got 0

<plh> http://w3c-test.org/framework/results/nav-timing-default/

Karen: We should let Boris follow up on the test_navigation_type_reload, as Firefox is the only UA failing.

plh: We should be good to submit very soon.
Resource Timing

JatindeR: http://dvcs.w3.org/hg/webperf/raw-file/tip/specs/ResourceTiming/Overview.html#performanceresourcetiming-methods
... The spec isn't clear on what to do when onresourcetimingbufferfull isn't defined. Should we have a circular buffer or throw things that are spilled over?
... Should we update the processing model to state that if the onresourcefull callback isn't defined we just don't add the resource to the primary buffer

Action Jatinder to update Resource Timing processing model to take care of the case that onresourcetimingbufferfull isn't defined

<trackbot> Created ACTION-100 - Update Resource Timing processing model to take care of the case that onresourcetimingbufferfull isn't defined [on Jatinder Mann - due 2012-04-18].
<scribe> Topic: User Timing

Jatinder: Updates measure method without any parameters to store the duration from navigationStart to the current time, rather than fetchStart to the current time
... Updated all examples to include sub-millisecond resolution
... ○ Throws an INVALID_ACCESS_ERR exception if either startMark or endMark argument, or both, have the same name as a PerformanceTiming attribute with a "DOMHighResTimeStamp" time value of 0.
... I deleted that text.
Performance Timeline

Jatinder: To make the behavior very clear, updated Section 4.4 startTime attribute to return 0, per WG discussion last week.
... I responded to the NAvigation Timing attributes in High Resolution Time in this mailing list topic: http://lists.w3.org/Archives/Public/public-web-perf/2012Apr/0027.html

Tony: I will ask James to respond to that topic.
High Resolution Time

Jatinder: I made the following high resolution time updates: updated DOMHighResTimeStamp definition to take into account durations, updated DOMHighResTimeStamp definition to recommend microsecond resolution on millisecond units, added a note saying if sub-millisecond not possible, should return millisecond, resolution in millisecond unit, added a note saying that time spent not fully active is not excluded from current time
TPAC 2012

plh: Do we want to go TPAC this November in France?
... If I recall, other than interacting with Cameron, we didn't work too much with other groups.

Arvind: I agree that we didn't interact with other groups and the overhead might be significant.

Tony: I agree, we should maybe just meet on the west coast.

Jatinder: I think there is value in our WG face to face meeting, but I do see that the overhead is high for France and we don't usually interact with other WGs as well.

Arvind: We should probably poll the mailing list, but it feels like a no for France. I

plh: I will follow up.

<plh> https://www.w3.org/2000/09/dbwg/details?group=45211&public=1&order=org

plh: Looks like we have decided not to do the face to face in France.

<plh> Future Web Technology, Baidu, Infraware, Qihoo 360, SK Telecom, Tencent
Page Visibility

Jatinder: I have made the following updates: updated Section 4.2 to move non-normative statements into notes and capitalized, MUST and MAY statements, added prerender as an optional state, added definition of the preview state, added an editorial note for Issue 8, added hook to HTML5 unloading processing model in the Page Visibility processing model