companies contributing to X [was: Re: Respository vandalism by root at ...fd.o]

[Matthew Garrett]

On Thu, Nov 25, 2010 at 09:23:38PM +0000, Alan Cox wrote:
> > but simply being more enthusiastic about accepting contributions doesn't 
> > seem like a great plan (compare the code quality of nouveau, intel and 
> > radeon to that of some of the out of tree drivers, for instance)
> 
> I think that is a little naïve. There is a difference between vendors
> attempting to use Xorg as a "dump and run" for crap code, and being a bit
> more relaxed about obscure drivers that are otherwise unmaintained.

I don't entirely agree. If people provide code review and the vendor 
maintainer's attitude is approximately "We're only willing to work with 
you if you accept our approach", I don't think that benefits us. It can 
be an opportunity for learning - I'm just not sure that it has been in 
the real world, so far.

> X is a bit "odd" in other ways - it's history has been rather closed at
> times which hasn't helped as it means there isn't a long standing large
> developer base.

That's certainly true. The small number of developers has been a 
longstanding issue, and the fact that companies can't really just pick 
up an existing developer makes all of this much harder.

> It consists (for much of the relevant stuff) of a very small number of
> very large and very complex drivers for insanely complex bits of
> hardware. That doesn't have the same scaling for newbies the kernel does
> where there are hundreds of random USB widgets you never knew you needed
> that make good starting points.
> 
> Maintaining the old Voodoo2 driver was a bit like minor kernel hacking. I
> can't even imagine how KeithP fits everything he needs to know for the
> intel drivers into his head.

The lack of documentation for various aspects of the server doesn't help 
either. I found X development far more intimidating than getting 
involved in the kernel.

-- 
Matthew Garrett | mjg59 at srcf.ucam.org