Re: [css3-images] Defining SVG paint servers as a CSS <image>

On Tue, Jun 21, 2011 at 5:53 PM, Tab Atkins Jr. <jackalmage@gmail.com>wrote:

> On Tue, Jun 21, 2011 at 1:56 PM, Rik Cabanier <cabanier@gmail.com> wrote:
> > On Tue, Jun 21, 2011 at 1:12 PM, Tab Atkins Jr. <jackalmage@gmail.com>
> > wrote:
> >> On Tue, Jun 21, 2011 at 12:19 PM, Rik Cabanier <cabanier@gmail.com>
> wrote:
> >> > It would even be better if there was a way that you could refer to a
> SVG
> >> > symbol instead of a paint server.
> >> > We're working on converting Flash content to HTML+SVG and if the
> >> > animation
> >> > is complex, we create many external SVG files.
> >> > The size of the SVG files is not a problem but having to do a http
> >> > request
> >> > for each one causes a lot of overhead.
> >> > If we could refer to symbols, we could put all our content in 1
> external
> >> > file which is much more efficient...
> >>
> >> That's theoretically doable by just targetting <svg> subelements
> >> directly, right?
> >
> > Correct but it requires JavaScript to make it work.
>
> Hmm, you miss my point.  The ability to refer to a single sub-<svg> of
> a larger SVG is part of the spec, right?  I think nobody does it, but
> I'll assume for now that it's something that can be fixed.


If it is part of the spec, I have not seen it.
Conceptually it would be identical to your proposal of "background:
url(doc.svg#mycooldrawing);" with 'mycooldrawing' defined as a symbol in
'doc.svg'.


> >> Alternately, it might be doable by a suitable interpretation of
> >> <pattern>.  By default, I'd treat a <pattern> as an infinite image
> >> constructed from positioning and tiling the contents.  Alternately, we
> >> could treat <pattern> as just its contents, and leave the tiling part
> >> to CSS.  That would be a bit more magical than I probably want to
> >> worry about, though.
> >
> > Yes, using patterns feels more like a hack.
> > I agree that for your proposal, patterns should either not tile or be
> > excluded as a paint server.
>
> I was actually saying the opposite - patterns should indeed tile,
> because that's how they're defined in SVG.  My definition would just
> have them tile into the canvas that CSS uses.  It was the non-tiling
> that was hacky and which I didn't want to do.


I was agreeing with this part of your proposal:
  >> we could treat <pattern> as just its contents, and leave the tiling
part to CSS.

The tiling as defined by SVG should be ignored and replaced by tiling as
defined by CSS.