[SCM TRANSITION] Re: Proposal: move Randr protocol and library to git
eich at suse.de
Tue Apr 18 23:43:14 PDT 2006
Alan Coopersmith writes:
> Egbert Eich wrote:
> > I think we are moving slightly of topic here - but: the ABI has been stable
> > (modulo glitches) for 6 years. This sucked. Numerous problems we are facing
> > today exist because of this paradigm. Decalring any ABI stable for such a long
> > period of time seems to be *insane*.
> Your insanity is Sun's business model, and X's reality. The libX11 ABI has
> been stable much longer than 6 years - we can still run 15 year old X binaries
> on Solaris today since while the libc & libX11 ABI's have grown, they haven't
> broken existing applications. Maintaining ABI stability is only insane when
> the ABI isn't designed to allow evolving it while maintaining compatibility -
> the XFree86 module ABI is at least partway there with the versioning, but
> clearly still has some holes that could have been better designed to allow
> change without breakage.
Alan, I was not talking about the Xlib ABI. That's a different story.
With all the versioning in place I don't see how the driver module ABI can
be extended in a same fashion as a lot of what is making things painful today
is inimately tied into DDX core functionality today.
We'd have to rewrite the entire DDX to fix this (which we will eventually
have to do anyway). But by the time we have reimplemented the old ABI in a
modernized environment nobody will care any more because everybody will have
moved the drivers to a new ABI that offers more flexibility.
I think we still should distinguish between ABIs to software were a lot
of binaries exist that may be impossible to port and more internal ABIs
like the driver modules where the code is either under our control or
under control of companies that are working closely together with us.
But I may be wrong and people can convince me otherwise.
More information about the xorg