SPAM: Re: GSoC CM collaboration

Graeme Gill graeme2 at
Mon Mar 3 07:18:48 PST 2008

Kai-Uwe Behrmann wrote:
> Am 04.03.08, 00:33 +1100 schrieb Graeme Gill:
>>Kai-Uwe Behrmann wrote:
>>>- a ICC profile container attached to window content
>>I don't thing that's a good approach, as it would necessitate
>>incorporating the CMM deeply into things. Better to pass down
>>a device link and leave the CMM higher up the stack. This
>>then allows an open choice of CMM's and linking approaches
>>(intents, smart linking, tuning etc), as well as supporting
>>device links.
> This brings the CMM into the applications and the rending backend.

You mean it splits it between the application or an upper level and
the rendering back end. Yes, that's the idea. A typical CMM
is really these two bits aggregated anyway, the linking code and the
pixel engine. The lower levels provide mechanism (pixel transformation
using the GPU), the upper levels provide policy (what profiles and how
they should be linked).

> For 
> what benefit? How does device link profile avoid the presence of a CMM 
> down the path?

By avoiding the need to link profiles in the low level code.

> As well, when a user wants to be able to select globaly a certain CMM and 
> his application opts in for colour management, as should be the default, 
> then there will be CMM specifics down the path in any case.

Sorry, I'm not understanding this sentence.

Graeme Gill.

More information about the xorg mailing list