> However, this has already been flamed to death from both sides, so until
> we have a viable alternative, live with it (as others have pointed out,
> GGI isn't yet ready, even if Linus were to allow it in the kernel).
I just recently thought about that (Dosemu died and took my console with
it).
Would not something like the following be enough to always be able to
recover?
1. Make the kernel know how to reset the video card to a known state. This
should essentially be a string of "output X to register Y"; a user land
application could configure it via some IOCTL.
2. Make the kernel _always_ react to some key sequence with this card
reset, and to another with a console switch. Even in raw mode. Maybe use
an obscure combination if you must.
This sounds like a fairly small change, yet one that might actually solve
the biggest problem.
(As to the userland configurator, I'd like to see something similar to the
OS/2 SVGA utility - in a V86 jail, do a video card init (and mode
switches), and simply record what the BIOS does. The people who wrote the
BIOS presumably knew the hardware. Do it as text, so you can fix it if
they didn't.)
MfG Kai