> > As soon as you have a port-specific driver where you could have
> > a machine-independent one, you immediately lose all of the benefits of
> > having it machine-independent.  Namely, the bug fixes and improvements
> > to one version no longer apply to the others, and there are fewer
> > people using & working on either so improvements come less quickly.
> 
> Of course.  I just can't see the benefits of a machine-independent, non-
> functional driver either. ;-)

The driver in question _is_ functional, at least at some basic level.
I use it on the Alpha all the time.  I'm writing this mail from a
386/20 with a 16550, dialed in at 38.4, and i see overflows
infrequently (i.e. on average, one silo overflow once every 15 or 20
minutes).



cgd