[Openicc] XICC specification draft (Xinerama vs. composite).

[Jim Gettys]

Xinerama as it currently exists has a number of shortcomings, but
many/most people find it useful.

By its nature, however, it is hiding multiple monitors as though they
are one: if they are not matched in characteristics, Xinerama present
major headaches for people interested in serious color.  It may be that
presuming if you are running xinerama that the monitors/flat panels are
a matched set is probably "good enough" for people who want to do that.

Another approach to the current Xinerama is possible once the new
composite extension deploys widely and our drivers work well for it
(right now, only certain X drivers can run composite effectively).
We've certainly been talking about the possibility of making xinerama
obsolete someday given composite.

For those of you not familiar with composite, here's a thumbnail sketch
of how it works:
	When composite is enabled, rendering no longer goes to the frame
	buffer(s); instead, the rendering occurs in offscreen pixmaps.

	An external application, called the compositing manager, becomes
	responsible for composing the real image in the real frame 
	buffer. (it gets told which windows have what pixels updated), 
	and the compositing manager may add eye candy of various sorts.

One could then have the compositing manager attempt to do a conversion
from, say, linear rgb to what the real frame buffer needs to drive the
particular monitor with color correction.  This conversion might be
different on different monitors that make up the unified "screen" seen
by applications. The issue then is whether the resolution of the
information in the pixels is sufficient; my intuition is it probably is
"most of the time", but probably not for serious applications.

But I'm not a color expert: you folks tell me ;-).
			Regards,
				- Jim