[Webkit-unassigned] [Bug 109884] Add CSS Property tracking to FeatureObserver. Creates new histogram for CSS Property usage data.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=109884





--- Comment #35 from Mike Lawther <mikelawther at chromium.org>  2013-03-26 23:35:21 PST ---
(In reply to comment #32)
> The feature seems very low value for all the bloat it introduces.

Hi Antti! I think it's very valuable to be tracking *at minimum* properties considered experimental (ie prefixed ones). Is the low value / bloat tradeoff you see due to currently tracking all properties?

Can we agree that tracking prefixed features (and their current/future unprefixed counterparts) as a minimum is worth having a mechanism in place? 

I would argue that in the future, any new CSS properties added should also be tracked as they are untested in The Real World and tracking their usage will be interesting. So setting things up so that an author is required to add them to the tracking makes sense. The switch statement added in this patch makes this a compile time check. 

Kassy's performance numbers show that the execution time for tracking *all* properties is in the noise.

> As I see it, the point of tracking these things is to figure out what APIs we can remove. Is there another benefit to tracking CSS property usage?

Removing is a black/white type decision. There are levels of gray that are useful as well :) It can also help inform any optimisation/tradeoffs we might make. eg if we want to try and do some gnarly optimisation to border-radius drawing, it'd be helpful to have an idea of how many pages/users would benefit. We've got data structures for 'rare' styles, but these aren't based on hard data from the field as far as I know. Tracking changes of properties over time can tell if we've got the makeup of these data structures right. Yeah, there are some obvious ones that are used everywhere - do you reckon we should make an exception for them?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.