New color conversion system

[Boudewijn Rempt]

On Thu, 20 Sep 2007, Cyrille Berger wrote:

> Yes. Anyway, acyclic graph isn't very difficult. The problem is to find the 
> correct heuristic to determine which path is the best. And the shortest is 
> not necesseraly the better. All colorspace have to be able to convert to/from 
> LAB16 (and to/from RGB16 for that matter), so the shortest solution will 
> allways be at a maximum of 2 color convertions.

I'm not sure what heuristics the kofilterchain code actually uses. Shortest
path sounds likely, though.

> And we had need to understand lcms code :/ (understanding the API is allready 
> quiet a big job for my little brain)

Your brain is pretty big! But the big issue, afaik is not even lcms itself, it's
the icc profile limitation. You once linked to a paper on the issue (or was that
Kai-Uwe) from 2006 with a proposal for adding float precision to icc profiles.
I never heard whether anything happened -- and then ctl got released with that
stupid license.

> How short are we ? We still have three monthes of alpha on the planning. And I 
> had say that finishing pigment is a top priority for 2.0.

That's true -- if you think that it'll be doable in, say, two months, why not
do it the right way. Maybe it's just my own nervousness, caused by having run
KDE4 for an hour...

Boudewijn