[Openicc] Linux CM ideology, was: meta data in test chart

[Kai-Uwe Behrmann]

Am 04.02.11, 11:18 -0000 schrieb Richard Hughes:
> On 4 February 2011 11:02, edmund ronald <edmundronald at gmail.com> wrote:
>>  When I want to write "Hello World", I don't want to have to choose a
>> font, a kerning, a color, a background color etc, allocate a buffer,
>> allocate a graphics port, image into the buffer and then copy onto the
>> port.
>
> This is why I think it's important to put this kind of thing in the
> toolkit. By pushing CM into projects like pixman and cairo, GTK gets
> it for free, and stuff magically works without any extra code from the
> programmer. The same can be done with Qt.

That would be really phantastic. Passing through of ICC profiles as meta 
data to backends and so on was discussed in long threads. Would be nice to 
see it actually happen. As well having Gtk and Qt handle ICC profiles 
during loading of images like PNGs would be great as well.

(However I do not see exactly what pixman relates to colour management.)

> Of course, that's platform specific, but I think that is a good thing.

Its not a good thing. As vendor of a Linux distribution you might argue 
this way. Application developers and even more end users want consitency, 
not have to learn three languages for the same thing. In that sense
it would be a terrible idea to make cairo behave different on osX, Linux 
or Windows platforms. Btw. thats completly against one major goal of 
cairo, to keep output consitent. So platform abstraction matters.

> By providing native high level hooks it's always easier to use than a
> generic "jack of all trades" solution that's trying to add a generic
> hook into the various layers of rendering code.

Who wants generic hooks?
Generic APIs are wanted, which provide abstraction from specific hooks.

> This is why colord and gnome-color-manager will never do buffer writes
> or pixel manipulation as they both will just feed high level data into
> the system and session to let the existing frameworks do the
> conversion in the right layers. This is one major point where I think
> myself and Kai-Uwe disagree :-)

Compiz provides hooks into the GPU during compositing and deploys 
Oyranos to provide fast desktop colour correction. Or the image_display 
code in Oyranos uses native CPU conversion through lcms, or 
an other CMM, to provide a very flexible approach. Yes, I agree, 
specific hooks are useful and widely deployed. They are on the plan
of OpenICC and Oyranos since years.


kind regards
Kai-Uwe Behrmann
-- 
developing for colour management 
www.behrmann.name + www.oyranos.org