[minutes] 20110330 Web Performance WG

Web Performance WG Teleconference #26 Agenda 2011-03-30
30 Mar 2011

See also: IRC log at http://www.w3.org/2011/03/30-webperf-irc 

Minutes
http://www.w3.org/2011/03/30-webperf-minutes.html

Attendees

Present 
TonyG, Nic Jansma, Karen Anderson, Jatinder Mann, Arvind, Jain, Zhiheng Wang, Christian 

Regrets 
Philippe Le Hegaret
Chair 
Arvind Jain 
Jason Weber

Scribe 
Jatinder Mann 

Contents
Topics
1. Feedback and discussion on expected failures for test case.
2. Feedback and discussion on test_timing_attributes_order.htm test case.
3. Feedback and discussion on updates to Resource Timing.

Summary of Action Items
[NEW] ACTION: Jatinder to add examples to Section 4.2. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action02]
[NEW] ACTION: Jatinder to update existing tests per Philip's first suggestion: test if window.performance and window.performance.timing exist, report failure if they don't, and abandon that. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action01]
[NEW] ACTION: Jatinder update Section 4.5 to make it clear when http header is not included. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action03]
________________________________________
<scribe> chair: Jason Weber
<scribe> agenda: 'http://lists.w3.org/Archives/Public/public-web-perf/2011Mar/0106.html'
<scribe> scribe: Jatinder Mann

1. Feedback and discussion on expected failures for test case.
<scribe> ACTION: Jatinder to update existing tests per Philip's first suggestion: test if window.performance and window.performance.timing exist, report failure if they don't, and abandon that. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action01]

<trackbot> Created ACTION-15 - Update existing tests per Philip's first suggestion: test if window.performance and window.performance.timing exist, report failure if they don't, and abandon that. [on Jatinder Mann - due 2011-04-06].

2. Feedback and discussion on test_timing_attributes_order.htm test case.

Tony: The link that I have doesn't discover the tests.

Will send out test case in mail and if no one has any concerns by end of the week, we will move it to the approved folder.

Previous comment regarding test_timing_attributes_order.htm test case.

3. Feedback and discussion on updates to Resource Timing.

Jatinder: Thanks Christian

... Let's review Section 4.2 Resources included in the PerformanceResourceTiming interface

Zhiheng: Should include examples in this section.

<scribe> ACTION: Jatinder to add examples to Section 4.2. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action02]

<trackbot> Created ACTION-16 - Add examples to Section 4.2. [on Jatinder Mann - due 2011-04-06].

Jatinder: Let's review Section 4.5 Cross-origin resources

Zhiheng: What about a hybrid solution that includes a meta tag for those that don't have access to http header?

Christian: Seems like such a meta tag would only be useful for html resources (iframes, etc).

<scribe> ACTION: Jatinder update Section 4.5 to make it clear when http header is not included. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action03]

<trackbot> Created ACTION-17 - Update Section 4.5 to make it clear when http header is not included. [on Jatinder Mann - due 2011-04-06].
We all agree that the meta tag isn't necessary, as html resources aren't as common.

Let's discuss Sigbjørn Vik email on Wednesday, March 30, 2011 2:55 AM

Jatinder: the goal of the type attribute is - Give web developers a way to bucket their resources. URL already gives the exact resource. URL is the URL requested, not the final URL.
... We should make the definition of type attribute clear: The Type attribute is the object that is initiating the resource download.

Nic: One of the goals of having the initiator attribute is to help web developers group their resources by common initiator type. For less commonly used initiator types, we should rely on INITIATOR_OTHER. The data is not lost, as the web developer can always look at the URL and ID attributes.
We all agree that the definition of attributes in Resource Timing should be explictly listed, instead of pointing to the Navigation Timing spec. This is because the definitions are slightly different (document vs. resource, navigation vs. fetch).

Nic: We want to remove loadEventStart and loadEventEnd. These are non-network related events. Developers could get this information already. These load events are also not application for all resources.We were also worried that having them here would confuse people that the loadEvent was inclusive of the time that the browser was loading and processing the resource. If there is no loadEvent fired or hooked into, would we have time stamps? Since developers c
draft.

Zhiheng: We should still include them. We have a similar situation in the Navigation Timing, but we kept them there. We should keep them here for consistency.

Karen: These are events coming from your previous page, in the Navigation Timing spec. But in the resource case, we do have access to this information.
... Referring to the unloadEvent.

Tony: I'm not sold on tying to DOM events.

Christian: I'm also not convinced we need to tie to DOM events.

Tony: id would also have the same arguments as loadEvents.

Nic: Without the loadEvents, developers can add up all the remaining times to see the total time to download the resources. Displaying is out of scope for this spec.

Zhiheng: I agree that we should remove this but add a comment in the spec. We can keep the current behavior, but solict feedback.

Nic: We've added the id attribute for ease of web developers to pull down information on resources that they have the id for.

Tony: One a simple page, one could pull the id. But in reality, pages are very complicated. The id can change.

Nic: We would keep the id as the moment the fetch starts.

We can make the id definition more clear to say "the id as stated at the moment the resource was requested."

All agree to keep the id.

Let's discuss the BufferSize requirement of 1000.

We all agree that the UA should recommend a 1000 entries rather than require. The UA can determine the buffersize - e.g., mobile browsing case. setResourceTimingBufferSize should return the maximum limit that UA supports.

Summary of Action Items

[NEW] ACTION: Jatinder to add examples to Section 4.2. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action02]

[NEW] ACTION: Jatinder to update existing tests per Philip's first suggestion: test if window.performance and window.performance.timing exist, report failure if they don't, and abandon that. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action01]

[NEW] ACTION: Jatinder update Section 4.5 to make it clear when http header is not included. [recorded in http://www.w3.org/2011/03/30-webperf-minutes.html#action03]

Received on Wednesday, 30 March 2011 22:59:33 UTC