modular -> monolithic

Russell Shaw rjshaw at netspace.net.au
Wed Jan 23 06:16:01 PST 2008


Egbert Eich wrote:
> Russell Shaw writes:
>  > Egbert Eich wrote:
>  > > Dave Airlie writes:
>  > >  > > This whole "partial" remonolithicalisation seems to not address the real
>  > >  > > issue, which is a wrong general mindset, and instead this equals
>  > >  > > sticking ones head in the sand and hoping that things magically go
>  > >  > > away.
>  > ...
>  > 
>  > The way it should've been done in the first place is to leave it all monolithic,
>  > but have a script to download only the parts of X and drivers required.
>  > 
>  > Only the top level would've needed a configure.ac
>  > 
>  > After a period of hacking had been done, it would be easy to download the
>  > rest of the monolith around what you'd already had, then do a complete build
>  > to see what's broken.
> 
> This is missing an important point: it was considered a feature
> to do out-of-band releases.
> Something users, distros and hardware vendors were pushing for hard.
> 
> Considering the nightmare we all had I still think it is a feature.

What is an out-of-band release?

In the monolithic model, drivers can still be released separately to X because
they're plugins that can be installed into users systems without recompiling.

Maybe some other parts of X could be made modules too.

> Currently we are seeing a lot of API breakages. Some people attribute these
> to the backlog of things that have piled up over the years.
> 
> If this is the case we should see less of these once the huge pile has been
> reduced. Then it may actually become rather simple again to make a driver
> that is API compatible for a range of server releases.
> 
> What do we do then? 
> 
> Remodularize?
> 
> Cheers,
> 	Egbert.
> 




More information about the xorg mailing list