Using HAL for X Server Config Properties?

[Jim Gettys]

There are a number of things going on here:

1) I think we are mostly agreed on probably using dbus to talk to the X
server to control configuration.  There isn't any particular Linux
dependency to dbus that I can see.

And configuration can get complicated, with many different input devices
and potentially more than one seat on a system.

2) Where to store configuration information: I think we mostly agree
that having this stored in some mystic X configuration file is a
mistake: the current situation is pretty broken.  We can argue about
whether this should get stored in a desktop specific configuration
system, something unique to X, or some yet-to-be-agreed on common
desktop configuration storage system.  But this is a separate issue, and
one for the desktop systems themselves to worry about (though so far, no
one seems to have successfully cracked this nut).

3) There is *a lot* of hardware related stuff involved in X.  We have
many different classes of input devices (mice, touchscreens, tablets,
joysticks, data gloves, etc.; last I looked, XInput had a dozen classes
or so, and people have been inventing new ones since it was defined),
which may require the startup of device drivers, loading of firmware,
restoration of calibration information after reboot, and possibly server
processes to handle remote input devices.  Ideally, a new device of a
manufacturer can get added without having to touch the X server (or wait
until a distro gets around to adding configuration GUI's for it).

Similarly, we have the advent of hotplug screens (e.g. PCIE), though
these have been around a while on cardbus and pcmcia.  Again, screens
should plug in and "just work".  I certainly agree with the sentiment
I've heard that plugging in a new screen, whether hot plug or not,
should result in a "working" configuration that uses that screen.  We
can argue about exactly what the default behavior should be.

I'm less familiar with HAL than D-BUS...  but certainly the above sounds
a lot like HAL's wiki page or Havoc's manifesto.  But I have no clue if
other platforms have been interested in picking up HAL: X serves many
different platforms, not just Linux.

4) Hotplug itself seems to vary between platforms; and I'm unaware of
how it works elsewhere at all (e.g. Mac OSX, Windows, *BSD).  Can people
enlighten us?

Certainly, on Linux, we want X to be a first good class citizen of the
desktop. And it isn't a requirement that everything run on all
platforms; we can expect platform maintainers to have to do some work, I
believe.  But I do think that the more we can share across platforms,
the happier everyone will be.

				Regards,
					- Jim





On Fri, 2005-10-07 at 10:56 -0400, David Zeuthen wrote:
> Hi,
> 
> I'm not sure it's really a good idea to involve HAL in the X.org server
> configuration process. I remember talking to krh and ajax about X and
> dynamic reconfiguration at the DDC a few months ago. The thinking was
> somewhat like this
> 
>  1. The X server starts up without configuring any hardware. It listens
>     on an IPC interface that is not the X protocol (it makes sense to
>     use a secure system message bus here, e.g. D-BUS) [1]
> 
>  2. When your display manager (e.g. gdm) and desktop session (e.g.
>     GNOME) starts it looks for X servers that are not yet configured
> 
>  3. A suitable unconfigured X server is configured. The settings (e.g.
>     resolution, color depth, multiple monitors etc.) is read from the
>     users settings (e.g. gconf). After a while the desktop session
>     can connect to the X server through the X protocol
> 
>  4. For on-the-fly reconfiguration the desktop session can connect
>     to the IPC D-BUS interface and reconfigure the X server.
> 
> The key point here is this: it is paramount that the X server doesn't
> read a system-wide configuration file. Actually, the departure from
> using a configuration file and instead provide an API that enables some
> program running in the desktop session to configure the X server
> provides this. This program might even fall back and try other options.
> 
> So, I think this is the ideal way to go about it. I don't think we want
> to include the HAL project in this :-)
> 
> Cheers,
> David
> 
> [1] : this interface would offer this like fingGraphicsCards(),
> findMonitorsByGraphicsCard() etc. though probably slightly more complex
> (also include input devices etc.)
> 
> 
> On Thu, 2005-10-06 at 10:48 -0700, Johnson, Charles F wrote:
> > Actually the X Server has to have quite a bit of knowledge of the
> > underlying hardware, particularlly on the display side if you  want to
> > take advantage of such things as HW Accelerated 3D.  The question of
> > multiple X servers is an interesting aspect that I hadn't considered.  
> > 
> > One question however, why would adding a x11 properly any different that
> > what has already been done in HAL for ALSA ??  
> > 
> > My real goal here is to investigate what should tbe the proper approach
> > so that the X server can make use of the hot plug infrastructure of the
> > underlying Linux system so that we can move away from manually
> > re-configuring the X server every time a hardware change is made.    
> > 
> > I've CC'd this onto the xorg mailing list also.
> > 
> > Thanks.
> > 
> > --Charles Johnson
> > Intel Corp.
> > charles.f.johnson at intel.com
> > 
> > >-----Original Message-----
> > >From: Pierre Ossman [mailto:drzeus-list at drzeus.cx]
> > >Sent: Thursday, October 06, 2005 6:47 AM
> > >To: Richard Hughes
> > >Cc: Johnson, Charles F; hal at lists.freedesktop.org
> > >Subject: Re: Using HAL for X Server Config Properties?
> > >
> > >Richard Hughes wrote:
> > >>> A paritial example could be:
> > >>>
> > >>> x11.type = 'pointer'
> > >>> x11.driver = 'mouse'
> > >>> x11.protocol='IMPS/2'
> > >>> x11.pointer.emulate3buttons='yes'
> > >>
> > >> IMO, makes sense. You could use a prober to populate this.
> > >>
> > >
> > >Is leaking information from the X server into hal such a good idea? An
> > X
> > >server does not have very strong ties to hardware and we could in fact
> > >be running several X servers.
> > >
> > >If this is pursued it should probably manifest itself as subnodes of
> > the
> > >hardware, one for each X server.
> > >
> > >Also, if you're a bit paranoid you probably don't want user specific
> > >information, like X server configurations, leaking into what is
> > >essentially a system specific database. ;)
> > >
> > >Rgds
> > >Pierre
> > _______________________________________________
> > hal mailing list
> > hal at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/hal
> 
> _______________________________________________
> xorg mailing list
> xorg at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/xorg