[minutes] Web Performance WG Teleconference #106 2013-04-17

Meeting Summary:



1.     Diagnostics and Error Logging

To ensure that the proposed Error Logging interface doesn't contain any security or privacy concerns, we've asked members to follow up with their security and privacy teams and raise any concerns with the interface with the working group.



2.     Resource Priorities

As a framework for discussion, we have uploaded a draft of the Resource Priorities spec: https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/ResourcePriorities/Overview.html. Seeing that the normative text is only a few paragraphs, we may want to eventually put this information in the HTML5 spec. The Web Perf WG should ensure that the feature text solves the performance problem before we add this to the HTML5 spec. We will continue the discussion in the mailing list.



3.     Resource Timing Updates

Resource Timing has been updated to only use well defined terms, instead of terms like "networking layer".



Microsoft will be reviewing and providing feedback on the Resource Timing test cases by next Wednesday's call.



4.     High Resolution Time L2

We have uploaded a draft for High Resolution Time Level 2, which now includes support for performance.now() in the worker context:  https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/HighResolutionTime2/Overview.html



5.     Agenda for next week's call

We will continue our discussion of Error Logging and Resource Priorities next week.




W3C Web Performance WG Teleconference #106 2013-04-17



IRC log: http://www.w3.org/2013/04/17-webperf-irc



Meeting Minutes: http://www.w3.org/2013/04/17-webperf-minutes.html



Attendees

James Simonsen, Jatinder Mann, Daniel Austin, Arvind Jain, Philippe Le Hegaret, Jason Weber


Scribe

Jatinder Mann



Agenda

1.     Diagnostics and Error Logging

2.     Resource Priorities

3.     Resource Timing updates

4.     High Resolution Time L2

5.     RequestAnimationFrame

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



Minutes:
Diagnostics and Error Logging
Jatinder: James had raised that considering the level of detailed information we are considering sharing, we should go through a security and privacy review of this feature. I plan on reviewing this feature with the IE security team in the next week or so and can get back to the working group.
James: Our security team has reviewed this and generally they are fine with this feature on a same origin. I need to follow up with the privacy team as well.
Arvind: I think we should definitely seperate out the historical and current interfaces, and probably in seperate specs.
Dan: If the historical interface has a higher privacy or security implications, we should seperate specs.
Jatinder: We may want to consider updating the spec to include JavaScript errors as well?
Dan: We may want to consider that in the future.
Jatinder: I propose we touch base on this topic after the security review.
Resource Priorities
Jatinder: I had put together a draft of the Resource Priorities spec as a framework for discussion:
https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/ResourcePriorities/Overview.html.
<plh> note that video has preload
<plh> which is a hint
Jatinder: This aligns closely with the discussion on the mailing list. Adding "defer" to any element or object capable of fetching a resource, will mean that the UA may choose to not download that resource until after all other resources without the attribute have started downloading.
... Adding "defer" to a script element will mean that in addition to not beginning the download, the UA may also not execute the script until after the page has been parsed.
... What I don't currently agree with is that style information or viewport visibility shouldn't impact whether the resource is downloaded. Determining intent is easiest for the developer and the developer should choose or not choose to add the defer tag.
... Also, I don't think any of these statements should be must clauses. They should be may.
<simonjam> load event blockers: http://www.whatwg.org/specs/web-apps/current-work/#delay-the-load-event
Dan: Considering the script element uses defer, I wonder if we should separate this to another attribute.
Jatinder: Seeing that script execution is already delay executed, it may make sense to combine these. But we may choose to use a new term like "delay".
James: What about changing the load event such that it's not impacted by the defer attribute resources.
Jatinder: I think we should think about this one some more. I wonder if we want to redefine the load event.
... If we do decide to define that this spec does not impact load, it may make sense to give the attribute a separate name, as real world sites may already use 'defer' with script and it will change behavior.
Dan: That's a good reason to change the name.
Arvind: Redefining load to push it out has the benefit of improving performance, as scripts can now execute sooner and not be delayed by deferred resources.
Jatinder: Let's treat changing the load event and using defer or a different name as open items that need further discussion.
... I don't think style information should impact whether a resource is delayed or not.
James: I agree with that as well.
Dan: Seems odd that something at layer 7, formatting, impacts something at layer 4, transport. I agree that style shouldn't impact.
Jatinder: There's also a performance impact on the browser to calculate that information at an earlier time.
... Let's agree to not include style information. What about viewport or above/below the fold? I think that the developer is the best person to determine that and they can do so by including or not including the tag.
Arvind: Is our goal to have a separate spec or should this be put into HTML5? Seems small enough to add to HTML5.
Jatinder: I think this spec is a good framework for discussion. If we feel that this is more appropriately placed in HTML5, we should do that.
Resource Timing updates
Jatinder: Spec has been updated to no longer refer to not well defined terms like "networking layer".
... Is adding a redirect-origin-clean flag really necessary?
James: We may not need the flag, but we may want to improve the redirectEnd definition.
Jatinder: Good feedback, I'll look into it.
High Resolution Time L2
Jatinder: We have posted a draft of the HRT L2 spec, which now includes support for performance.now() in the worker context. Spec is here: https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/HighResolutionTime2/Overview.html
... Does everyone agree that the start of creation of a shared worker is step 1 of the worker processing model, here: http://www.w3.org/TR/workers/#processing-model
... "Create a separate parallel execution environment (i.e. a separate thread or process or equivalent construct), and run the rest of these steps asynchronously in that context."
James: I agree, makes sense.
plh: I can get help get that editorial change made.
RequestAnimationFrame
plh: There is a discussion on whether or not rAF calls should be made for display:none or below the fold animations. Seeing that the spec is in last call, we need to decide how whether we want to take that feedback or not.