W3C home > Mailing lists > Public > public-web-perf@w3.org > July 2011

[minutes] 20110706 Web Performance WG Teleconference #40

From: Jatinder Mann <jmann@microsoft.com>
Date: Wed, 6 Jul 2011 22:28:02 +0000
To: "public-web-perf@w3.org" <public-web-perf@w3.org>
Message-ID: <EE4C13A1D11CFA49A58343DE361B0B04068B2197@TK5EX14MBXC252.redmond.corp.microsoft.com>
Meeting Summary:



1.      Web Performance WG to meet at TPAC 2011 in Santa Clara first week of November 2011

The WG agreed to meet at TPAC 2011 in Santa Clara, CA on November 3rd. This will be a good opportunity to discuss status of all the specs in the Web Performance WG and network with other working groups present. Philippe has an action item to book the room.



2.      First half of the meeting was cancelled this week

Due to vacations, we had cancelled the first half of the conference call meeting reserved for discussions on Navigation, Resource and User Timing. We will commence the conference call next week.



3.      Feedback and discussion on the setImmediate specification

There are a few discussions happening on the mailing list regarding feedback and clarifications on the current setImmediate design, including clamping, allowing the UA to delay the event, script based schedulers, frequency and power consumption benefits of this API. In this call, pros and cons were discussed for not clamping the setImmediate API. It was decided to continue the discussion on the mailing list.



4.      Discuss requestAnimationFrame open issues

a.       The following open issues of the requestAnimationFrame spec were discussed:


1.       ISSUE-1: Scheduling processing model needs to be more tightly defined

James will be working on updating the spec to define the processing model in terms of an HTML5 task queue, so that animation frame callback scheduling participates in the browser event loop in a well defined manner.



2.       ISSUE-2: Callback time parameter needs definition

James will be working on updating the spec to be more clear that callbacks should match the refresh rate of the display.


3.       ISSUE-3: Animation frame times should be monotonically increasing

The spec needs a section to clarify that the animation frame times should be monotonically increasing, similar to what was stated in the Timing specs: http://www.w3.org/TR/2011/WD-resource-timing-20110524/#monotonic-clock. The spec should also include window.animationStartTime as a standardized way to get the start time for all animations in a given refresh interval, allowing them to remain in sync with one another.


4.       ISSUE-4: We perhaps should support an element parameter to requestAnimationFrame()

James will work on defining an optional element parameter to this API.


5.       ISSUE-5: Expected callback rates should be documented
There were a few callback throttling algorithms discussed for the case where the page is no longer visible, including no callbacks, exponential backoffs and a throttled callback. The WG agrees that this portion of the spec should be normative, rather than optional, to ensure most interoperability and least compatibility issues. Both Google and Microsoft implementations use the no callbacks algorithm, and they have not seen any issues so far. Cameron will follow up to see if Mozilla also agrees with this approach.


6.       ISSUE-6: Spec needs to clarify expected behavior for duplicate calls of the same callback

This issue has been closed and no spec changes were required.


7.       ISSUE-7: FrameRequestCallback interface should be designated as Callback=FunctionOnly
This issue has been closed and no spec changes were required.


b.       Once the above issues have been closed, we can target taking the spec to Last Call. Per Charter, we had agreed to do so by August 2011. Let's target first week of August to take this spec to Last Call.



6.      Discuss Page Visibility open issues

a.       Page Visibility Updates

The Page Visibility spec has been updated to include a Vendor Prefixed section and PAGE_PRERENDERED state has been removed.



b.       Creating a test suite for Page Visibility

Karen Anderson is out sick, however, will be working on providing the Page Visibility test cases shortly.



c.        Last Call for Page Visibility

The spec is still on schedule to enter Last Call on July 13th. The Last Call period will last from July 13 - August 10th.



We had the following action items from this meeting:

1.       Cameron McCormack and James Robinson: Make requestAnimationFrame spec updates for next week.

2.       Karen Anderson: Add Page Visibility test cases.



Detailed Notes:



Web Perf Teleconference #40 7/06/2011



IRC log: http://www.w3.org/2011/07/06-webperf-irc


Meeting Minutes: http://www.w3.org/2011/07/06-webperf-minutes.html



Attendees

Present for Navigation Timing, Resource Timing and User Timing (4-5PM EST/1-2PM PST)
Meeting cancelled this week.


Present for Page Visibility, Efficient Script Yielding, Display Paint Notifications (4-5PM EST/2-3PM PST)
Arvind Jain, Jason Weber, Jatinder Mann, Philippe Le Hegaret, James Robinson, Cameron McCormack



Scribe

Jatinder Mann



Contents

Agenda
1.       Discuss feedback on the setImmediate specification
2.       Discuss status on the Page Visibility specification

3.       Discuss status on the RequestAnimationFrame open issues
4.       Any other business




--------------------------------------------------------------------------------
Discuss feedback on the setImmediate specification.

Jatinder: Any particular feedback on the setImmediate spec we want to discuss first?

James: Per my mailing list question, I had some concerns regarding the clamping.

Jason: I had just sent a response to you on the mailing list prior to the call starting.
... you had some questions on my historical account, what in particular are you referring to?

James: Questions about the historical account on when clamping was introduced.

https://bugzilla.mozilla.org/show_bug.cgi?id=123273

Jason: We are excited about setImmediate, we want to have an API to that helps with power efficiency, including on intel and arm chips.

James: Do you think this will let us remove the clamp?

Jason: Yes, absolutely.
... We would love to get to a pattern where the clamp is removed for sites that use these apis, allowing for better efficiency.
... this may also help Chrome in Speedstep, as the 4ms clamp is probably impacting your implementation. You should be able to see this in the Power Consumption blog post we wrote a few months ago.I can respond on the mail.

James: If you can send specific examples that'll be great.

Jason: Any other questions on setImmediate or can we move on to requestAnimationFrame?
Discuss feedback on the requestAnimationFrame specification.

<plh> http://www.w3.org/2010/webperf/track/issues/raised

Jatinder: Per discussion in last week's call, the requestAnimationFrame has 7 open issues in it that are blocking the spec from going to last call. Most of these issues, as discussed, do have relative consensus in the group and just need to be specified in the spec. Cameron, James, if you are overbooked and need any help with updating the spec, I am more than willing to help you. Please let me know.

James: Issues 6 and 7 can be closed, as there are no spec changes required.

<plh> issue-6?

<trackbot> ISSUE-6 -- (duplicate callbacks): Spec needs to clarify expected behavior for duplicate calls of the same callback [Request Animation Frame] -- raised

<trackbot> http://www.w3.org/2010/webperf/track/issues/6

<plh> issue-7?

<trackbot> ISSUE-7 -- FrameRequestCallback interface should be designated as Callback=FunctionOnly -- raised

<trackbot> http://www.w3.org/2010/webperf/track/issues/7

James: Issues 2 and 4 may need more time to write the text here.

Jason: Issue one (scheduling callbacks) could become a compat issue if it isn't well defined in the spec.

Jatinder: We discussed in the past that we can either do no callbacks, an expontial backoff or a throttled callback. We discussed whether to make this optional or normative. I recommend we make this a normative requirement to ensure implementations are interoperable.

James: Agreed.
... I prefer no callbacks, as that is most efficient.

Jason: Agree that we shouldn't do callbacks.

James: Cameron, can you see if there are any concerns on your end with not issuing any callbacks?

Jatinder: If I recall, some concerns with no callbacks was that the animation wouldn't be drawn. However, if you do time based animation and you use the timestamp parameter, the callback shouldn't matter, as the animation would be drawn at the proper place when visible.

James: Agreed.

Cameron: I will respond to the mailing list with this option and see if Boris or others have any concerns.

James: I plan on closing most of the editorial issues soon.

Jatinder: Once we have resolution on the issues, we can move towards taking this spec to last call.

Cameron: What does the charter suggest?

Jatinder: Charter suggests August 2011 to enter Last Call.

Cameron: This seem reasonable.
Discuss feedback on the Page Visibility specification.

Jatinder: Per action items from last week, the Page Visibility spec has been updated to include a section on Vendor Prefixes and to remove the PAGE_PRERENDER state. Per WG decision, we plan on taking this spec to last call on 7/13.

Arvind: Has there been any feedback on the spec?

Jatinder: None on the mailing list.

Arvind: Great.
TPAC 2011

Jatinder: TPAC 2011 will be a great opportunity for the Web Performance WG to touch base on the current status of the specs, and to network with other groups in the area. Philippe, per your request last week, can we arrange something?

Philippe: What does works best? One day or two?

Arvind: 1 day. Indifferent which day.

Jason: Agreed, one day.

Philippe: Let's look at the Thursday.

Jatinder: Thanks. If no other issues, let's see everyone next week. Thanks!
Received on Wednesday, 6 July 2011 22:28:44 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:04:31 UTC