> That's a good start, but we really need to abstract the VDAC (or
> ramdac) controller out of that code, instead of replicating it 3 times
> -- or 4 times, if you count i860 boards with 24-bit colour via 3 VDACs.

Sensible, but IMO the abstraction should be flexible enough to support
things like:
	
* It should be the card's responsibility to set the colormap
  using the correct function (i.e. at vertical blank time), once it's
  notified that the colormap is dirty. If you don't do this, you'll
  get shearing. Or there should at least be a hook into the bt* code
  notifying it of the interrupt.

* In the PX driver for instance, we have to use the VDAC ID on the
  lower/mid-end 3D cards to figure out the number of planes in the
  framebugger. There's no other easy way of doing this (if you probe
  by using the stamp you'll get deliberate corruption on the screen),
  i.e. useful macros should be exported from the VDAC header files.


Andy.