Re: minutes, Tokyo 2013 SVG F2F Day 1, 3 June 2013

Robert O'Callahan wrote:
>         heycam: Same.
>         ... Seems like this will suck.
>
>
> Actually we do layerize some SVG transforms for GPU compositing, as of
> quite recently. There are some bugs blocking the performance
> improvements from being visible, but they're getting fixed.

My mistake.  (Although have we merged transform="" and the 'transform' 
property entirely?)

> I think SVG transforms should cause an isolated group.

My "seems like this will suck" above was in response to any use of 2D 
transforms in SVG causing a stacking context.  I feel like transforms 
are used in SVG much more freely, and wouldn't cause authors to think 
that it would cause the same restrictions as opacity, etc.

>         krit: We make them isolated groups, and I think FF does too.
>         ... Also masks and clips are stacking contexts.
>
>
> Opacity has to create a stacking context, and therefore has to create an
> isolated group.
>
> If people want blending to break out of stacking contexts or other
> isolated groups somehow --- let's see the use-cases for that and see if
> there's any reasonable way to support them. Getting filter Background
> and blending/compositing working inside isolated groups seems like
> enough work for now!

I guess that's reasonable.  If filters with BackgroundImage are the only 
current SVG feature (ignoring newer things like blending) that would be 
impacted by any new stacking contexts existing (as transforms could), 
and given BackgroundImage is not widely implemented, we can start with 
transforms creating a stacking context (<=> isolated group) to begin 
with, and then relax that later if we can or need to.

Received on Monday, 3 June 2013 23:55:16 UTC