- From: Tab Atkins Jr. <jackalmage@gmail.com>
- Date: Mon, 29 Apr 2013 15:41:40 -0700
- To: Blake Kaplan <mrbkap@gmail.com>
- Cc: Dimitri Glazkov <dglazkov@chromium.org>, "Edward O'Connor" <eoconnor@apple.com>, "public-webapps@w3.org" <public-webapps@w3.org>
On Mon, Apr 29, 2013 at 3:14 PM, Blake Kaplan <mrbkap@gmail.com> wrote: > On Mon, Apr 29, 2013 at 3:00 PM, Dimitri Glazkov <dglazkov@chromium.org> wrote: >> As far as implementation complexity, <content select>, distribution >> APIs, and <shadow> are trivial, compared to the event handling and >> representation of the composed tree. Hoping to alleviate this, I wrote >> all event-related handling as imperatively as I could. > > While I agree that specifying <content select>, distribution APIs, and > <shadow> is relatively simpler than some of the other parts of the > spec, I'm pretty worried right now about the performance of those > features, especially for dynamic changes. I just found out, as well, > that Mozilla's XBL implementation of these features worked correctly > in the static case, but is completely wrong in the dynamic case, so we > don't have any real data on how slow doing that stuff correctly is. Distribution-as-specified, with its reprojection and such, will definitely incur some performance penalties. However, we've worked through the alternatives, and there aren't any that aren't simple terrible. Literally every thing we tried to find that doesn't involve reprojection instead involves losing the ability to arbitrarily compose, and when you *do* compose, you need to write some truly terrible selectors to make things work right. ~TJ
Received on Monday, 29 April 2013 22:42:27 UTC