Handling of driver protocol extensions in cvs
jamey at minilop.net
Tue Jan 31 01:10:26 PST 2006
On Tue, Jan 31, 2006 at 09:26:01AM +0200, Aivils Stoss wrote:
> On Otrdiena, 31. Janvāris 2006 01:35, Adam Jackson wrote:
> > While the latter idea sounds fine, Aivils' proposal to me sounded like
> > "protocols are hard and I don't want to define my API so let's just not
> > worry about it", which is completely not the way to go.
> Why *nix ioctl fuctions works without API around 30 years. But under
> X i should write MyCoolDriverSetGibberingOn() - 100 lines of code at
> least, sane is 200 lines. These 200 lines do the job:
> Gibbering = 1
> Nice, nice, very nice!
Solving this "problem" was one of the design goals of XCB:
Now you just have to write a few lines of XML, and the client-side
protocol binding is generated for you automatically. In principle we can
generate the server-side binding too; that just hasn't been implemented
yet. So theoretically the only code you should have to write is the
server-side part that actually does something.
Adopting a completely untyped ioctl-style interface will not make things
better: in my opinion even GLX is fairly painful for this reason.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 189 bytes
Desc: Digital signature
More information about the xorg