Tobias Jakobs wrote:
If you're wondering what to do about gradients with multiple stops and stuff like that, if worst comes to worst we could always resort to using the computed values for a lookup using feDisplacementMap.
Can you please post an example image? I'd like to know how it would look like.
I've posted a few examples on-line: http://home.hccnet.nl/th.v.d.gronde/gradients/
The idea is documented on, feel free to comment, improve, etc.: http://wiki.inkscape.org/wiki/index.php/Advanced_Gradients (Sorry, not nicely formatted math, yet.)
However, as you can see on that page the rendering is not (yet) very pretty. With the linear gradient it's not too bad, but the conical gradient is quite bad. In addition, there is a reason I rendered those with Batik, Inkscape's filters are not yet able to properly render them. On the other hand I guess these make nice test cases :)
The quality of the renderings can definitely be improved by using higher bit-depths, higher quality implementations of the filters and increasing the complexity (for example by segmenting the region). But of course there may still be better ways to create a conical gradient (or any other gradient) that I haven't found yet.