[minutes] 20101013 Web Performance

Available at:
http://www.w3.org/2010/10/13-webperf-minutes.html

Web Performance Working Group Teleconference
13 Oct 2010
See also: IRC Log
http://www.w3.org/2010/10/13-webperf-irc

Attendees
Present
AndersonQuach, NicJansma, TonyG, Zhiheng, ArvindJain
Regrets
JasonWeber
Chair
ArvindJain
Scribe
AndersonQuach
Contents

  *   Topics<http://www.w3.org/2010/10/13-webperf-minutes.html#agenda>

1.       Consensus on the Navigation Timing definitions

2.       Feedback on the timing diagram

3.       Feedback on the processing model, if we include navigationStart, this may inadvertently reveal information about redirectStart and redirectEnd in cross-domain scenarios

4.       Time permitting, Resource Timing and User Timing

5.       Any other business

________________________________

NicJansma: navigationStart, what about navigations without unloads?

Zhiheng: this should line up with the HTML5 spec
... Any concern with the definition fetch start?

AndersonQuach: Looks good, as long as we do not enforce undefined phases.

TonyG: Sounds good.

NicJansma: domContentLoaded split to domContentLoadedEventStart, domContentLoadEventEnd?

TonyG: Jonas brought that up, to be consistent with events.

AndersonQuach: should we added it?

TonyG: Works for me.

Zhiheng: yep.

NicJansma: navigationStart is stated after the redirect phase, should it start before redirect phase to be closer to user committing the navigation?

TonyG: the navigationStart moves is implemented in WebKit as in front of the redirect phase.

NicJansma: IE as well.
... navigationStart moved in front of redirectStart can expose the redirect phase.

TonyG: All of the folks presented on the onsite respect privacy concerns, and review with browser security team and summarize the issue on an email thread.

NicJansma: navigationStart and prompt for unload should happen prior to redirect.

Zhiheng: it is possible to have unload phase in the HTML5 spec?

TonyG: Can you share that reference?

<Zhiheng> http://dev.w3.org/html5/spec/history.html#navigating-across-documents

Zhiheng: Note in the beginning of 4.5 stating the lifetime.

NicJansma: Add one more sentence in step 1, they miss that.

Zhiheng: Good idea.

NicJansma: Step 3 in the processing model, there may be a typo. -- 3. If the current document and the previous document are not from the same origin set both unloadEventStart and unloadEventEnd to 0 then go to step 5. Otherwise record unloadEventStart as the time immediately before the unload event fires.
... What are the differences between step 5 and 6.

Zhiheng: if http
... before app cache lookup, and step 6 is not an http get.
... should these be merged?

NicJansma: Yes, it would help.
... Step 15, sub-bullet 4.. add redirectStart and redirectEnd in the exclusion list.

Zhiheng: navigationStart should point to the start of the redirect, not consistent with the diagram.

NicJansma: this section will be dependent on our joint security review.

AndersonQuach: it is our understanding that no new window object is created during a redirect. Tony is that similar to WebKit.

TonyG: Yes, no new window object is created for 300s with a subsequent 200.

Zhiheng: Graph is non-normative, itself does not have to be conformance benchmark.
... you're correct we do not have to enforce the gaps/ undefined phases.

NicJansma: Are there any other business with Navigation Timing Spec

<scribe> chair: JasonWeber