Re: Comments on SVG 2 Gradients

On Thu, Dec 8, 2011 at 11:27 AM, Tavmjong Bah <tavmjong@free.fr> wrote:
> On Thu, 2011-12-08 at 05:47 +0000, Cyril Concolato wrote:
>> The next item to be discussed on the SVG 2 Requirements list is
>> Gradients. I’ve a made a small page
>> (http://www.w3.org/Graphics/SVG/WG/wiki/SVG2GradientsComments)
>> describing some comments that I would like to discuss, possibly if we
>> have time during the next telcon. Let me know if you have comments.
>
>        I've seen the same problem you point out with some of the drawings I've
> made. It shows up when one has a large change in color as your black and
> white example shows. The problem can also be seen with the existing SVG
> linear and radial gradients. I've been planning on adding tensor meshes
> to my version of Inkscape to see if moving the tensor handles can
> minimize the effect.
>
>        You point out the problem can be seen to some extent with my color
> conical gradient example. This is the exact same problem as one has with
> linear and radial gradients. I verified this when I first saw the
> problem and wanted to know if it was a bug in the Cairo implementation
> or if it was inherent to linear/radial/Coons patch mesh gradients.
>
>        The paper you refer to is interesting, especially the mesh optimization
> part (from Inkscape's point of view). However, I wonder in practice if
> the drawings would appear much different if the meshes were simply Coons
> Patch meshes. For the most part, all the meshes have small changes in
> color where the problem you've pointed out would not be so apparent.
>
>        For me, the question becomes: do we want to specify the more
> complicated Ferguson meshes (or something similar) in the SVG spec or
> stay with the more simpler (and "standard") Coons Patch/Tensor Patch
> meshes and let the authoring software take care of any problems (like
> Illustrator does when exporting to PDF).

The most important consideration from my perspective is what can be
efficiently implemented in browsers and in GPUs.  A slow gradient will
kill performance on the whole page.  So, whatever the prettiest thing
that can still be efficiently implemented is what we should go with.

~TJ

Received on Thursday, 8 December 2011 21:54:13 UTC