[minutes] 20110511 Web Performance WG Teleconference #32

Thank you for meeting on the call today. The meeting summary is shown below:



1.      Discussed updates made to the Resource Timing spec

We agreed to the following changes made to the spec:



a.       Default buffer size decreased to 150 from 1000

b.       Remove fetchEnd and made secureConnectionStart optional

c.        Remove getResourceTimingsByType method

d.       Monotonic Clock section



2.      Discussed additional updates to be made to the Resource Timing spec

The group agreed to making the following changes to the spec:



a.       Update responseEnd definition

We discussed updating the definition of responseEnd to remove the last sentence, as we now only show responseEnd and not fetchEnd:

"If the last non-redirected fetch of the resource is not the same origin as the current document,responseEnd must return zero unless the Timing-Allow-Origin HTTP response header rules apply."



b.       Update Section 4.2 Examples clear

Nic to update the first example in this section to make it clear what type of cache item is cleared.



c.        Update Section 4.1 to include SVG

Nic to update Section 4.1 to include reference to SVG.


d.       Update Section 4.3 and Processing Model to refer to Timing-Allow-Origin

We agreed that for clarity the spec should refer to the Timing-Allow-Origin cross-domain behavior in both the attribute definitions and processing model. Jatinder to make these changes.



e.       Update Section 1 Introduction

Jatinder to update this section providing background into the purpose of this API and give an example of using this API to measure resource timing information.


3.      Take Resource Timing to First Public Working Draft and Last Call

The WG has decided to release the Resource Timing spec as a First Public Working Draft on 5/25, in two weeks. Due to W3C moratorium, we will not be able to do so sooner. We are also tentatively setting 5/25 as the same date to enter Last Call, provided we do not find any major issues between now and 5/25. Members of the WG are encouraged to circulate the spec for feedback from interested parties.


4.      Discuss updates to the User Timing spec

Tony and James to provide a proposal to update the marks() and measures() API by next conference call. The WG can evaluate whether the new proposal provides substantial benefit compared to the current specification as soon as we have a proposal for review. Jatinder to update the User Timing processing model section as soon as we have finalized on the marks() and measures() API.



5.      Discuss updates to the Page Visibility spec.

Reviewed the following updates made to the Page Visibility spec:



a.       Agreed to not support onvisibilitychange

We agreed that addEventListener provides the same functionality as onvisibilitychange, with the additional benefit of registering multiple listeners. Considering that this spec defines visibility on the page and not the element, not being able attach a listener from markup is not a concern.



b.       Agreed that PAGE_OTHER is not required.

Considering User Agents can easily just prefix visibility states, the PAGE_OTHER state is not required.



c.        Agreed to make PAGE_PRERENDER and PAGE_PREVIEW states optional, but keep PAGE_VISIBLE and PAGE_HIDDEN as required

Considering some UA or Operating Systems may not support a prerender or preview state, these states need to be defined as optional.



d.       Agreed to use DOMString constants instead of unsigned short constants

There was some discussion whether we should provide constants at all, however, the value of allowing web developers to inspect for capabilities via IDEs seems valuable. String constants are preferred over short constants.



e.       Hidden to not include fully obscured non-minimized window case

We agreed that it would be difficult to implement the fully obscured non-minimized window case as being hidden. It's not yet clear whether the spec should be specific or generic on this point, as this scenario may not impact all platforms (e.g., mobile devices may not have this issue).



We had the following action items from this meeting:

1.       Tony Gentilcore and James Simonsen: to follow up with a proposal to make the measure API more coherent with the marks concept.

2.       Jatinder Mann: Update responseEnd definition in Resource Timing

3.       Nic Jansma: Update Section 4.2 Examples clear in Resource Timing

4.       Nic Jansma: Update Section 4.1 to include SVG in Resource Timing

5.       Jatinder Mann: Update Section 4.3 and Processing Model to refer to Timing-Allow-Origin in Resource Timing

6.       Jatinder Mann: Update Section 1 Introduction in Resource Timing

7.       Jatinder Mann: Update Page Visibility spec with proposed changes.



The meeting minutes are given below:



Web Perf Teleconference 5/11/2011



IRC log: http://www.w3.org/2011/05/11-webperf-irc



Meeting Minutes: http://www.w3.org/2011/05/11-webperf-minutes.html



Attendees

Present

Cameron McCormack, Jatinder Mann, James Robinson, Zhiheng Wang, Arvind Jain, Tony Gentilcore, James Simonsen, Christian Biesinger, Nic Jansma, Karen Anderson, Ben Weigand, Philippe Le Hegaret



Meeting Chair

Jatinder Mann



Scribe

Jatinder Mann



Agenda

1.       Feedback and discussion on Resource Timing.

2.       Feedback and discussion on User Timing.

1.       Feedback and discussion on Page Visibility.

2.       Feedback and discussion on requestAnimationFrame.

3.       Any other business.



Topics

Feedback and discussion on Resource Timing Processing Model.

Feedback and discussion on updates to Page Visibility spec.

Feedback and discussion on updates to AnimationRequestFrame spec.



We agree that to decrease default buffer size to 150 from 1000



Made the change to remove fetchEnd and made secureConnectionStart optional



Made the change to remove getResourceTimingsByType method



Suggest removing the last sentence in the definition of responseEnd, as we are now only showing responseEnd and not fetchEnd:



Remove this sentence from responseEnd: "If the last non-redirected fetch of the resource is not the same origin as the current document,responseEnd must return zero unless the Timing-Allow-Origin HTTP response header rules apply."



We all agree with this change.



Zhiheng: In 4.2 "Downloadable resources that are retrieved from the user agent's cache must be included in the PerformanceResourceTiming interface."

... this seems to contradict the first example.



Nic: The first example can be made a bit more clear what kind of cache item is not being included.



Action Nic to update the first example in Section 4.2 to make the cache item clear.



<trackbot> Created ACTION-27 - Update the first example in Section 4.2 to make the cache item clear. [on Nic Jansma - due 2011-05-18].



Action Nic to add SVG to Section 4.1



<trackbot> Created ACTION-28 - Add SVG to Section 4.1 [on Nic Jansma - due 2011-05-18].



Zhiheng: * In 4.3, We need more clarification about the accessing model about redirectStart and redirectEnd.



Action Jatinder to update Section 4.3 and Processing Model to refer to Timing-Allow-Origin header section.



<trackbot> Created ACTION-29 - Update Section 4.3 and Processing Model to refer to Timing-Allow-Origin header section. [on Jatinder Mann - due 2011-05-18].



The WG agrees that we should spend the next two weeks looking into the spec and preparing to enter Last Call. We should forward the spec to all interested parties for feedback.



move to agenda 2



Tony and James to follow up with a proposal to make the measure API more coherent with the marks concept.



<Christian> can we rename +1.650.253.aacc to [Google], like [Microsoft]?



<Christian> thx



<BWeigand> Can +1.503.712.aaaa be marked as [Intel]



<plh> zakim won't remember those btw. it needs.



Let's target FPWD for the Resource Timing spec for two weeks from now.



<plh> s/needs further intervention for that/



move to agenda 3



<jamesr> cool



1. We are all leaning towards document.visible. Jatinder to analyze data to see which of the following names to use for PageVisibility: window.visible, document.visible, window.pageVisible and document.pageVisible.



We are okay with not supporting onvisibilitychange.



Making prerender and preview states as optional.



Using DOMString constants instead of unsigned short constants



scribe: Need constants to allow web developers to inspect for capability

... webIDL spec, http://dev.w3.org/2006/webapi/WebIDL/#idl-constants, allows for DOMString constants



Don't believe we need to require PAGE_OTHER. UA can always add prefixed state.