[minutes] Web Performance WG Teleconference #113 2013-07-10

Meeting Summary:



1.     TPAC Registration

Registration for TPAC 2013 is open and the registration cost will increase after Oct 18. Please register: https://www.w3.org/2002/09/wbs/35125/TPAC2013/. The working group will send out a detailed agenda, but in general we will work on discussing the new specs in our charter and meet with other working groups to discuss some of our new ideas.



2.     Alternative Conference Call Time

To accommodate folks in Australia and New Zealand, we'll be suggesting an alternative conference call time slot that we can use from time to time when folks from those regions are interested in joining the calls. Philippe will follow up with options.



3.     Preflight JS Injection and Spotlight

Alois will be sharing a proposed document of the Preflight JS Injection concept, so that we can review the idea with security folks and ensure there are no major issues here.



There was a lot of discussion on the Spotlight project, including technical details that will need to be hashed out if we are to pursue this project. Dan will be sending out a detailed proposal so we can review and provide feedback.




W3C Web Performance WG Teleconference #113 2013-07-10





IRC log: http://www.w3.org/2013/07/10-webperf-irc



Meeting Minutes: http://www.w3.org/2013/07/10-webperf-minutes.html



Attendees

Philippe Le Hegaret, Jatinder Mann, Rob Dickenson, Alois Reitbauer, Dan Austin


Scribe

Jatinder Mann



Agenda

1.     TPAC

2.     Conference Call Times

3.     Diagnostics & Preflight JS Injection

4.     Spotlight Project

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



Minutes:



Agenda topics: TPAC, conference call times, Resource Timing question, preflight JS injection, and spotlight
TPAC

<plh> http://www.w3.org/2013/11/TPAC/#GroupSchedule2

<plh> https://www.w3.org/2002/09/wbs/35125/TPAC2013/

Alois: Can we have more information on TPAC and when the group should meet?

Plh: After Oct 18, the registration cost increases. I don't believe there is a deadline on registration.

<plh> we're meeting on Thursday and Friday

Alois: We should come up with an agenda for the working group.

<plh> http://lists.w3.org/Archives/Public/public-html-admin/2013Jul/0001.html

Dan: There were details from the HTML WG mailing list on getting Visas from China. That will be useful details for this groups as well.

https://www.w3.org/2002/09/wbs/35125/TPAC2013/?login#wbsq15

Jatinder: As for agenda, we should come up more details as we get closer. But two things we should include are make progress on our current chartered specs and also speak with other groups.

Plh: Seeing that Beacon, Resource Priorities, prefetch/prerender are interesting to the HTML WG, and we may even put it into the HTML spec, we should setup some time to meet with the HTML working group.
... Interestingly, I have heard people talk about taking dependencies on Beacon API, and interest in pursuing Element Visibility.

Jatinder: Let's take an action to find an alternative time slot that works for all of our attendees. In particular, we need an alternative slot that works for Australia and New Zealand.
Prefetch JS Injection

Alois: The idea we discussed was from navigation error logging where the server was down. The idea is that the HTTP would point to a JS that would run in any cases when the page is down. This way we will get real time availability.

Dan: What about security concerns of injecting JavaScript?

Alois: Analytics already inject JavaScript, so this isn't new. For HTTPS pages, you already can't change the content.

Jatinder: Alois why don't you put together a spec proposal? We can use that document to share this concept around to see if others feel that this may be reasonable from a security point of view.
Spotlight Project

Dan: I have sent information on the Spotlight. I'm hoping that the working group and others can look into making a public database of performance database of websites on different browsers. Make this data available to everyone. This will be similar to HTTP archive, but it'll be from a performance point of view. The goal here would be that we'll put the onus on site owners to improve their performance. We would also want to provide guidance to h[CUT]
... their performance. E.g., if you look at the waterfall data of the top 50 sites, you'll see that there is a lot of interesting performance data. With this data, we can push these folks to really improve their real world performance.
... I want this group to get behind this and W3C really has the credibility to make this happen.
... I have spoken to many companies, and they have really found this interesting, including from Google, Amazon, and others. This was also on the list of topics that Philippe covered with the AC. Philippe any comments from them?

<plh> --> http://www.w3.org/2013/Talks/0610-performance/ my slides from the AC meeting

Plh: The AC reps didn't seem very interested in this.

Jatinder: The IE team has spent the last six year develop one of the most sophisticated perf labs, so we have some insight here. For example, the sites here will change daily. There will be very large variability on a daily basis as web pages change their content. So I'm not sure how much data can be gathered. At the same time, the top five news sites or travel sites though they do similar things, have very different performance characteristics[CUT]
... would start to compare these sites as apples to apples, even though they are quite different. Another aspect that I want to understand is the machines used for testing. For example, in the IE Perf Lab we have main pool of machines that we pulled from the Dell Factory floor - simply changing a hard drive in one machine will change its performance characteristics. We really need to understand the details here, especially if this is public and us[CUT]

Dan: I think we're okay with variability, because if some Ad takes 20seconds longer, we would want to show that. I don't think variability is the big concern here.

Alois: What about logins? There are a lot of pages that will require logging in. I wonder who will be paying these costs.

Jatinder: What level of data would we show? It would be useful to help users see causes and correlations.

Dan: We would definitely want to provide as much data as possible, as well as possibly the HAR file.
... You can think of this as YSlow scores or something similar to WebPerfTest.org.

Jatinder: Please send some more information so we can review this to understand the challenges