[minutes] Web Performance WG Teleconference #99 2013-02-27

Meeting Summary:



1.     Beacon API

The working group spent most of the meeting discussing the Beacon API spec draft that Alois and I shared with the conference call attendees. Upon updating the draft based on the feedback from the call, I will email the draft to the mailing list for additional feedback.



2.     Agenda for the next two weeks

The working group plans to discuss diagnostics APIs in next week's conference call and prerender feature in the conference call the week after. Please come prepared to discuss those topics.




W3C Web Performance WG Teleconference #99 2013-02-27



IRC log: http://www.w3.org/2013/02/27-webperf-irc



Meeting Minutes: http://www.w3.org/2013/02/27-webperf-minutes.html



Attendees

Jatinder Mann, James Simonsen, Philippe Le Hegaret, Arvind Jain, Ganesh Rao, Rob Dickenson


Scribe

Jatinder Mann



Agenda

1.     Discuss Test Cases issues

2.     Feedback on Existing Specifications

3.     Feedback on New Specifications

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



Minutes:
Test Cases

http://w3c-test.org/webperf/tests/submission/Intel/resource-timing/

Jatinder: I noticed that Resource Timing test cases still don't work in IE. I'll follow up in mail.
Beacon API

https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/Beacon/Overview.html

https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/Beacon/Overview.html

beacon("POST", "/log", analyticsData);

James: I want to make sure that the spec is very clear that just because you are building data on the unload, we don't want to make people think that we're going to wait minutes to make this call.
... What if we have 5GBs of data? Do we want to set limits?

<plh> I added a link to the draft from http://www.w3.org/wiki/Web_Performance/Publications

Jatinder: Sounds reasonable that we put in some limits. Don't want the browser to wait forever.

Arvind: Should we keep the domain cookie or not for this response? I'm not sure what we should do here.

James: What about XHRs?

Daniel: If the beacon is ad related I don't expect it to impact cookies.

Arvind: We should make sure that the spec text doesn't say gauranteed, it should say best effort.

Jatinder: There were some considerations that I hadn't included in this spec. What if the browser is shutdown? Should we persist this information on disk and attempt to send it out next time? What if the browser gets an error in the response, does it attempt to send the data out again?

James: I think we should stick with the current, simple model where we attempt to send it once asynchrnously. If there is demand to increase our attempts or do some persisting work, we can consider that in L2 or based on feedback.

Jatinder: Okay, I'll keep the approach simple and move away from gauranteeing this will work to trying a best effort attempt.

Arvind: I want to make sure that Firefox reviews this spec as well, as they had privacy concerns with the Ping API. I can see that this API is different from Ping and tries to accomplish something different, but we should make sure there are no privacy concerns.

Jatinder: I'll update the spec based on today's feedback and send mail to the mailing list for additional feedback. I'll make sure Boris is CC'd.
Next Week's Agenda

Jatinder: Should we cover prerender or diagnostics next week?

Arvind: Let's cover prerender two weeks from now, as I will be out next week.

Alois: We can cover diagnostics next week. I'll follow up with you offline.