Kernel support for graphics cards

[Bojan Smojver]

On Mon, 2006-01-09 at 16:20 +0900, Carsten Haitzler wrote:

> so you have 1. a currently unsolvale techincal problem

If Linux/*BSD/Solaris get a modern graphical subsystem that makes sure
only one component is in charge of things that are currently done from
multiple components with little coordination, I think
Nvidia/ATI/<insert_vendor_here> will have no problem making relevant
kernel modules. In fact, they do it now (ship kernel modules for their
cards). I would also like them to open up the specs and the code, but in
the end it is their decision.

In other words, if they follow the new direction, they get to sell more
cards to us. I think they'll quite like that.

> 2. a viewpoint problem
> that both x and kernel coders agree on to overcome,

That's exactly why I (as a user) would like to see a page on
freedesktop.org that gives direction in relation to these matters (and I
think Dave is putting something up). A "plan" of sorts, so that everyone
knows where (roughly) things are supposed to go. Then the real technical
discussion can start between the people that actually know how to do
this (which then conveniently excludes me :-).

I quite enjoyed reading Keith's paper, because it was obvious that
things are going to get a lot better. I found plenty of references on
freedesktop.org about many things from the paper and their status. But I
didn't find much about this (and I had plenty of lockups on my machines
caused by lack of it), so I asked.

> 3. a LOT of work to do to
> desing a system thyat will work for everyone who needs it, 4. actually having
> to implement it all.

Well, with my huge kernel and X hacking experience of exactly 0 (zero)
lines of code, I have to admit I'm not overly worried ;-) Seriously, if
Microsoft could do it, I'm sure various kernel hackers
(Linux/*BSD/Solaris) and X developers can do it too.

> i'd say you are pushing poo uphill here :)

Hey, story of my life ;-)

-- 
Bojan