X server 1.10 development cycle

[Luc Verhaegen]

On Tue, Sep 14, 2010 at 05:55:41PM +0300, Mart Raudsepp wrote:
> On Fri, 2010-09-10 at 11:41 -0700, Keith Packard wrote:
> > Merging Drivers:
> > 
> > I'm also still interested to hear how we can de-modularize the system
> > to some extent where it would be useful. Merging the protocol headers
> > into a single package would be a good step, but it would require
> > someone to step up and offer to take charge of doing release
> > management for the whole thing.
> > 
> > I think Peter Hutterer is almost convinced that bringing the input
> > drivers into the server would help let us fix the API/ABI issues in
> > that interface faster.
> > 
> > Video drivers seem a bit harder, they're much larger, of course, and
> > we'd really need a dedicated 'subsystem maintainer' to make that work
> > inside the server. For drivers using KMS, I think moving them into the
> > core server would let us refactor them to eliminate the duplicate mode
> > setting code.
> > 
> > Anyone want to volunteer to have "their" driver get merged into the
> > server for 1.10?
> 
> I haven't heard any actual technical benefits in doing so, whereas there
> are many drawbacks in doing so.
> So I doubt any volunteering from the xf86-video-geode crowd, for
> instance. We want to support installing a newer driver version against
> older server versions, and that is quite easy to maintain if the driver
> is separate - even across video ABI changes, making a recompile of
> driver the only necessity.
> In addition it would seem to make it considerably harder to test GIT
> versions of the driver against the system installed xserver, and also
> various other drawbacks (packaging, etc).
> 
> Where can I read why merging (video) drivers would be a good idea in any
> form?
> 
> 
> Regards,
> Mart Raudsepp

Keith apparently just wants to kill off the linux desktop forgood.

The driverstacks, as they are laid out today are trying to abstract in 
the wrong place, where the interfaces are and should be moving the most 
(between the different pieces of the same driver stack instead of 
against the external APIs), and care is not and cannot be taken properly 
to keep even that back and forwards compatible.

The only place where this is today attempted is the xserver, and now 
this should be removed too?

So, we are working on tying xserver, mesa and the kernel 1-1 versionwise 
together. We just haven't managed to sell this entirely yet, but work is 
being done in getting people to swallow that.

Afterwards people will have to hear that "this is how it should be, as 
there is no alternative"... But noone will add "... as we killed off the 
alternatives".

In any case, this will make users extremely happy when they hit a bug in 
either graphics driver, acpi or wireless. You cannot either go and fix 
the bug yourself, as you cannot go and upgrade just one of them, or you 
have to update everything. When then the opengl or x libs change...

And who here has ever installed anything where both suspend/resume, 
wireless and graphics works to a satisfactory level out of the box?

What we then get is throw-away installations, where 1 image gets created 
for 1 set of hardware, and it never gets upgraded, it gets replaced 
completely. No more desktop systems where you can apt-get upgrade 
anything, as you might just as well install from scratch.

Preloads (where a team of people at SUSE or canonical get some hardware, 
spend many manyears making sure everything works perfectly together) or 
embedded style devices (where some company pays a lot of money to get a 
limited amount of devices working for very specialised purposes, with a 
team dedicated on keeping it up to date) are then the only uses for free 
software on the desktop. Of course, servers will always be possible in 
such a world (yes, redhat was right all along).

Guess how many people will be willing to install a windows or buy an 
apple machine all of a sudden, because that's the only way that they 
will get any work done?

So Keith, if you keep this up, at least be honest about it, and state 
openly and directly that you believe that the linux desktop is dead, and 
own up that you are about to drive the last nail into its coffin.

Luc Verhaegen.