[minutes] Web Performance WG Teleconference #116 2013-08-07

Meeting Summary:



1.     RequestAnimationFrame Updates

Per feedback, we may want to consider updating requestAnimationFrame spec to be generic enough to allow browsers to not trigger a callback when the browser knows that the animation cannot be visible. We can refine this statement in the V2 spec when we have potentially better element visibility definitions. There is a thread tracking this discussion: http://lists.w3.org/Archives/Public/public-web-perf/2013Aug/0016.html.



2.     Resource Timing Spec and Test Cases

The onresourcetimingbufferfull is now an EventHandler. We will also make Performance inherit EventTarget, so you can listen for the resourcetimingbufferfull using addEventListener().



The Resource and User Timing test cases submitted by Intel, Google, and Microsoft show failures in some test cases in the latest IE and Chrome, test harness timeouts, and duplicate test warnings. We need to determine if the test case bugs have been fixed, and if so, are these issues due to browser bugs, test case bugs, or spec bugs. We shouldn't move the spec forward to PR until we have resolved these issues.



3.     Beacon to FPWD

The working group has agreed to publish the beacon spec as a FPWD draft. Over the course of the next few weeks, we will also work on publishing Resource Priorities, Navigation Error, and Resource Error specs. These specs still have a few editorial clean up issues.



4.     Next Conference Call

Due to a number of vacations, we will schedule our next conference call for August 21.






W3C Web Performance WG Teleconference #116 2013-08-07





IRC log: http://www.w3.org/2013/08/07-webperf-irc



Meeting Minutes: http://www.w3.org/2013/08/07-webperf-minutes.html



Attendees

Jatinder Mann, Philippe Le Hegaret, Alois Reitbauer, Dan Austin, Arvind Jain


Scribe

Jatinder Mann



Agenda

1.     RequestAnimationFrame

2.     Resource Timing

3.     JS Preinjection

4.     Specs to FPWD

5.     Web Perf Site Design

6.     Next Meeting

--------------------------------------------------------------------------------



Minutes:
RequestAnimationFrame
Plh: Current text: Whenever a Document's hidden attribute is false and the animation frame request callback list is not empty, the user agent MUST regularly queue a task that samples all animations for that Document's top-level browsing context.
... Proposed text: Whenever a Document's hidden attribute is false and the animation frame request callback list is not empty, the user agent MUST regularly queue a task that samples all animations for that Document's top-level browsing context unless the user agent is able to determine that the impacted elements are hidden.
Jatinder: I can think of several cases where nothing is displayed
... in the past, we don't have a clear meaning from page visibility on element visibility
... but leaving it open would be fine by me if it resolved Boris' concern
... this is generic enough for v1 or we can revisit in v2
Resource Timing
Jatinder: I've updated onresourcetimingbufferfull to be an EventHandler
Plh: Once this change has been reviewed we can move to LC and get more review. We can then push from LC to PR directly.
Jatinder: I've looked at the test cases, and there are some failures in both IE and Chrome, test harness time outs, and some duplicate tests. We should verify if the test cases have been fixed, and if so, if the remaining issues are browser bugs, test case bugs, or spec bugs prior to moving Resource/User Timing to PR.
JS Preinjection
Jatinder: I wanted to talk about the goals: The monitoring JavaScript becomes a single point of failure for the page if it does not load, Modification of page source can result in unwanted side effects, The monitoring JavaScript will not be loaded when the page cannot be loaded (e.g. because of a DNS error), It does not allow to collect performance data for non-HTML request (e.g. images, pdf downloads)
... Only #3 seems like the one we had discussed, I want to understand what these other use cases are for
Alois: The first one is general case of JavaScript can impact page load. The second one is that depending on when you add the script to the page, you can get side effects.
Jatinder: I'd like to see some examples of those side effects.
Alois: I can provide that.
Jatinder: The second use case looks like it's more related to RUM monitoring, where you are detouring and instrumenting sites. Site or library owners can add User Timing directly in their code, and avoid the observer effect of having to detour code; technically, the detouring will slow down their page load.
Jatinder: The third bullet was the main goal from my point of view for this feature. As for the fourth bullet, Resource Timing does give you the image data.
Alois: The PDF download isn't associated with a Document, so that one might be missed.
Jatinder: We may want to consider that with Resource Timing L2, where downloads referred from a page are included in the timeline. It's not linked to the page load, but may be an interesting network perf data point.
Jatinder: I still need to have our security team review this spec. Once I have that review, I can provide more feedback.
Specs to FPWD

Jatinder: I have updated Beacon API spec with the working group feedback. I'd like to move this spec to FPWD.
Plh: I agree.
Dan: I agree as well.
Jatinder: I will be making updates to Resource Priorities, Navigation and Resource Errors next to get them ready for FPWD.
Web Perf Site Design
Jatinder: Plh, can we add the browser test case passing table to the site? Is this automated or manual?
Plh: The test case table is currently manually generated. We can look into doing that. If there are any updates you want to make to the site, do suggest or feel free to propose changes.
Next Meeting
Plh: I will be out next week on vacation.
Alois: So will I.
Jatinder: Let's schedule the next web perf meeting for two weeks from now.