From: James Simmons <jsimmons@edgeglobal.com>
  To  : ggi-develop@eskimo.com
  Date: Tue, 2 Feb 1999 10:08:33 -0500 (EST)

Re: The battles.

> > The result would be mixed user-space SW and accels,
> > introducing context-switching that reduces the speed of
> > the draw-line primitive.

What? The patch I/Fabrice wrote doesn't work that way. The reason behind
the what accels do I support ioctl was to avoid this mess. If you send
packets that the kernel doesn't support it will just ignore them. The
kernel should prevent hardware from crashing a system but it should
prevent someone programming bad. So if you send a accel thats not in
hardware it will be ignored. In this case you will the know the program
bug right away since you will not see the results you want. 

> > If the primitive had a kernel-side SW implementation,
> > a context-switch would possibly have been avoided.
> 
> If the primitive had a kernel-side SW implementation, no-one would use the
> driver, and this would be a non-issue :)

Right it would be slow and you do suffer from the context switch if you
use a software emulated accel in the kernel. Plus that would be HUGE
bloat.