Defining Seats

Chris Tyler chris at tylers.info
Wed Jul 18 12:44:19 PDT 2007


On Tue Jul 10 12:14:34 PDT 2007 Daniel Stone wrote:
>And since there's no real sensible way to share a
>GPU between X servers (although this becomes more feasible with
>kernel-based modesetting and AIGLX, but still), this basically means one
>video card per seat, which makes life easier.


There is a lot of interest in doing multiseat on multiple-output video cards:
- low-end solutions - one card for economy
- laptops - internal and external monitors as separate seats
- high-density multiseat - 2-4 outputs per card, 5-7 cards per system

Right now this is usually done with Xephyr, Xnest, or perhaps Xgl (?) on
top of a multi-screen base server; eventually we'll need better ways of
doing this.

It would a shame to completely preclude the possibility of a
multiseat-on-one-GPU configuration (or make it hard to configure).
Imagine the win if you could add multiseat capability to a system just
by plugging a monitor into a secondary video output and adding a USB
keyboard and mouse.

Perhaps for now we need a general framework in GPM for nested X servers;
if it's toggled on, a hotplugged monitor would be added to the
bottom-level server as a new screen, then a new top-level (Xephyr or
equivalent) server would be started to provide a new seat.

If this can't happen automagically (which is admittedly hairy) then we
need a manual way of configuring it.

--
Chris Tyler



More information about the hal mailing list