[Openicc] VMs and display correcotions

[Chris Murphy]

On Mar 9, 2011, at 1:34 PM, Graeme Gill wrote:

> Chris Murphy wrote:
>> I think that's a good way to do it. Display profile in common between host and guest,
>> vcgt ignored by guest, honored by host. There are other combinations possible but I
>> think that's the most sensible, off hand.
> 
> I don't see how that works in general. The host is meant to be providing
> an invisible virtual environment, so it is the component that should be
> doing nothing, leaving the guest with the illusion that it is talking to
> the real hardware, and doing the color management. Expecting the guest to
> be VM aware and change its behaviour is defeating the purpose of a VM,
> whereas the host knows it is a host OS.

vcgt honored by the guest OS would change a virtual LUT, which would cause a localized alteration of the VM guest window on the host and it certainly would end up with incorrect results because that localized region now has in effect two vcgt's applied.

There is a purpose behind LUT curves. So long as the result is arrived at, we don't actually need the VM to honor vcgt. YOu'd only need guests having control of display LUTs in a case where a host has NOT applied vcgt which is an even worse potential.

As for 128 guests all performance display compensation, for who knows how many remotes with VNC, I think that it's untenable. It's not scalable. There's other infrastructure needed.




Chris Murphy