Re: Minutes, SVG WG Brussels f2f day 4 (Monday)

Chris Lilley wrote:
 > anthony: Hi frequency color changes do not work as well (from the
 > author and from Anthony)

Do you have examples of this that I could play with?

 > shepazu: We are discussoin whether to make it a paint server, or
 > whether you apply via an effect
 > ...
 > <AlexD> I think you are putting the cart before the horse. It's a nice
 > graphical demo, however it would be good to first see mainstream
 > authoring tools support it; and more importantly fully evaluate the
 > run-time efficiency of the techniques. If the diffusion curve
 > algorithms are n^2 then forget it. Is there any IP also?

I'm not that worried about the complexity, so far it seems to be quite 
fast (although it's always hard to pin down the complexity of a 
numerical method exactly). And assuming that with n you mean the number 
of pixels I seriously doubt it would be n^2. But I'll definitely verify 
that.

What I am a bit worried about is the non-locality of diffusion curves 
though. This ties in to the comment above about whether to make it a 
paint server or an effect, as that would be one way to at least 
constrain the output a bit. But even then you can't really render 
different parts of the fill independently. This could probably be 
mitigated by either again going back to something like a multigrid 
method or by changing the algorithm to play nice with tiling.

Received on Thursday, 3 June 2010 08:01:45 UTC