[minutes] 20110427 Web Performance WG Teleconference #30

Web Perf Teleconference 4/27/2011

IRC log: http://www.w3.org/2011/04/27-webperf-irc 

Meeting Minutes: http://www.w3.org/2011/04/27-webperf-minutes.html 

Attendees
Present 
Jatinder_Mann, Karen_Anderson, Nic_Jansma, Jason_Weber, TonyG, JamesS, Christian, Zhiheng, Kyle Simpson 

Regrets 
Philippe

Scribe 
Jatinder Mann 

Contents
Topics 
Feedback and discussion on test case updates. 
Feedback and discussion on Monotonic Clock text. 
Feedback and discussion on Resource Timing Processing Model. 
Discussing the issue of including ID and Type 
Feedback and discussion on updates to Resource Timing and User Timing 
Discussing Page Visibility spec 

Summary of Action Items 

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

<Jatinder> scribe: Jatinder Mann

<Jatinder> ScribeNick: JMann

<Jatinder> move to next agenda

Feedback and discussion on test case updates.
<Jatinder> Tony: The test case that you had mentioned in the mailing list looks good to me.

<Jatinder> Zhiheng: I will send email with the test cases that I have submitted so everyone can review.

<Jatinder> Nic: I'm a bit concerned with the persistent connection test case - this isn't a normative requirement of the spec.

<Jatinder> Tony: What is the requirement in the spec that we are testing?

<Jatinder> Zhiheng: domainLookup and others should be set to fetchStart...

<Jatinder> Tony: This is a hard one.

<Jatinder> Nic: I'm concerned we can't force the browser to get into this state. The browser could chose to open a new connection even if it supports persistent connections.

<Jatinder> Zhiheng: I have tested persistent connection with most browsers (except safari) and they appear to be working as expected.

<Jatinder> Tony: This is not normative, we shouldn't be testing conformance behavior for non-normative.

<Jatinder> Nic: There is a requirement if a browser supports persistent connections, then a persistent connection should be used.

<Jatinder> Tony: On the other hand, it means we have made a valuable API, as there is no other way to tell whether a persistent connection was used.

<Jatinder> Karen: Only thing I can think of is flooding the network layer with lots of connections and checking it that way. Not sure if it will get us in this state.

<Jatinder> Zhiheng: I agree that we want to leave this case at this time. Let's look at other low hanging fruit.

<Jatinder> move to next agenda

Feedback and discussion on Monotonic Clock text.
<Jatinder> We agreed that the new text is good, however, there was concern whether this should be made normative or not.

<Jatinder> http://lists.w3.org/Archives/Public/public-web-perf/2011Apr/0072.html

<Jatinder> Zhiheng: I have concerns with the last sentence: The user agent must record the system time at the beginning of the navigation and define subsequent timing attributes in terms of a monotonic

<Jatinder> ...clock measuring time elapsed from the beginning of the navigation.

<Jatinder> Jason: The goal is to reduce clock skew. Having a start point will reduce skew?

<Jatinder> Is the issue about using a relative monotonically increasing clock?

<Jatinder> Jason: There are two pieces: using a monotonic clock and using a baseline. Without using a baseline, you can get a skew. In one browser, you may get inconsistent timing between sites. If not normative or well defined baseline, tests cross-browsers will get different results.

<Jatinder> Jatinder: Zhiheng, is your concern that if a better algorithm comes out in the future and we make this current text normative, browsers will not be able to innovate without being out of conformance?

<Jatinder> Zhiheng: Yes.

<Jatinder> Jason: If hardware comes with a better clocking algorithm, we can update the normative text in the specification to reflect that.

<Jatinder> Kyle: Monotonic to me means that the clock won't be negative and it would deal this issue.

<Jatinder> James: I'm a fan with monotonic, but not a fan of UTC.

<Jatinder> Jason: Can you clarify the divergence here?

<Jatinder> Jason: Let's talk through a specific example. If you run some tests on Chrome today, sometimes you get a negative time. This is probably due to offsets and what not. But for navigation time, we want to ensure that the time is baselined to the navigationStart and the timing is always increasing and never negative. If you use the system clock, instead of your own internal clock, the system clock will always be monotonically increasing and will be aligned with

<Jatinder> ...clock.

<Jatinder> James: Oh, that makes sense. I thought you were referring to the clock on the Taskbar.

<Jatinder> Jason: No, referring to the system clock.

<Jatinder> James: That sounds reasonable.

<Jatinder> Karen: The system clock is the system tick count.

<Jatinder> Jason: Does the wording sound better with this context?

<Jatinder> James: This makes more sense to me in this context.

<Jatinder> http://lists.w3.org/Archives/Public/public-web-perf/2011Apr/0072.html

<Jatinder> http://lists.w3.org/Archives/Public/public-web-perf/2011Apr/0072.html

<Jatinder> Let's discuss the wording further on the mailing list, considering we all believe in principal with the text and making this a normative statement.

<Jatinder> move to next agenda

Feedback and discussion on Resource Timing Processing Model.
<Jatinder> move to next agenda

<Jatinder> a. http://www.w3.org/mid/BANLkTinAF%252Bo9XigG14=uo2Hz-W0dHhKPCw@mail.gmail.com;list=public-web-perf

Discussing the issue of including ID and Type
<Jatinder> Jason: ID and Type will more uniquely identify dynamically fetched resources.

<Jatinder> ...- Minimum information you need is the timing data and identifying the resource that is associated with the timing.

<Jatinder> Tony: Worried about having all this additional data just for IDs.

<jamesr> Zakim: who is on the call?

<jamesr> heycam: is there a README for zakim?

<heycam> there is somewhere...

<heycam> probably he just wants a comma instead of colon

<jamesr> why does the web client tab-complete with a colon, then?

<heycam> http://www.w3.org/2001/12/zakim-irc-bot.html

<jamesr> hmm, pretty sure the last guy is me

<heycam> jamesr, that's a good question :)

<Jatinder> move to next agenda

Feedback and discussion on updates to Resource Timing and User Timing
<Jatinder> http://dvcs.w3.org/hg/webperf/raw-file/tip/specs/PageVisibility/Overview.html

Discussing Page Visibility spec
<Jatinder> Kyle: Privacy concerns. What if a page annoys the user if they are no longer viewing the page.

<Jatinder> ...the event may make this easier.

<Jatinder> Jason: Popups are something the user agent already suppress.

<Jatinder> Jatinder: A user agent may consider providing an option to not show the page visibility state.

<Jatinder> Jatinder: Let's aim to have all three specs available on the http://dvcs.w3.org/hg/webperf/raw-file/tip/specs/ site in the next few days, so we can begin discussing these specs on the mailing list. This will give everyone some time to form an opinion.

<Jatinder> ...We can discuss particular contentious issues on the call.

<Jatinder> Cameron: As per my mailing list question, are there any constraints on the spec name based on the new charter?

<Jatinder> Jason: No restrictions. The charter is worded to be vague enough.

<heycam> trackbot, url?

<trackbot> Sorry, heycam, I don't understand 'trackbot, url?'. Please refer to http://www.w3.org/2005/06/tracker/irc for help

<heycam> trackbot, status

<heycam> are we using tracker for issue tracking? (is what I'm trying to determine by asking trackbot. :))

<Jatinder> Cameron: As per my mailing list question, are there any constraints on the spec name based on the new charter?

<Jatinder> Jason: No restrictions.

Received on Wednesday, 27 April 2011 22:12:33 UTC