Plans for dynamically re-configurable X server

[Egbert Eich]

Alex Deucher writes:
 > Has anyone started working on this?  Is there an API in the works? 
 > Perhaps something like Xv attributes, only for the Xserver itself
 > rather than overlay specific?  The last time I brought this up was
 > prior to xorg, and as I recall, some of the powers that be at the
 > time, said there was something in the works, but I haven't heard
 > anything since.  xrandr covers some of the functionality, but it's not

Right. I had started on this. My idea was to make those the driver
options configurable that are desireable to change at run time and
do away with the limitations of both the xf86misc and xf86vidmod
extension.

However then the notion came up that this is nothing that should
be done thru an extension as it is local to the machine the server
is running on and therefore doesn't need to be network transparent.

Even xrandr shouldn't be there as the Xserver should just sit on
top of a generic driver layer which has its own configuration sceme
which doesn't need to be controlled thru the Xserver.

 > fully functional, and it'd be nice to support more chip specific stuff
 > like enabling/adjusting/disabling tv-out, dualhead, power management,
 > re-run DDC, and other tweakable settings.  Most drivers could handle

Some of these things should probably be configurable. 
DDC should really be rerun automatically when the display 
is hot-plugged.
We need to work on a detection for that.

 > it with fairly minimal work, but we need a system to interact with the
 > X server.   I guess one of the issues in memory management.  Ian is
 > working on a DRI memory manager and I think Alan H was working on one
 > as well.  While the current system isn't great, I don't think it
 > should be an impediment to adding configuration support.  All we would
 > really need is a current configuration validation function that would
 > enable/disable memory dependant features like the DRI, Xv, certains
 > depths/modes, etc. if there wasn't enough memory in the new
 > configuration.  Has anyone done any work on this?  Any thoughts? 
 > Ideas?  Let's kick off some discussion.
 > 

I don't know how desireable it would be to tune those memory parameters.
Certainly only the power user would profit from this. I think we would
ask too much of the average desktop user if we expect him to tune those
to meet his needs. 


Egbert.