> 	Would you have a reference to details on why UDI is such a bad
> 	idea? Is it an overspecified cumbersom model? designed to be
> 	optimised for no one OS, so performs poorly in all?

I'm trying to find the mail... basically, the overall idea of UDI is misplaced
now. It was a great idea 10 years or more ago (when I was one of the four
or five principle architects of the Solaris DDI/DKI- one of the others was
involved in UDI). Now we (overall, all of Unix) would lose far more than we
would gain by forcing a marshalling generic interface set for differet driver
classes.

A lot different platform device driver commonality can be achieved by a
careful design. Improvements in compiler technology and the reduction of the
actual number of really different hardware platforms has made the rubber glove
model less necessary...

I really had a long piece of mail about this to Mike Smith somewhere- I can't
find it. Basically, the gist was "Nice idea (a long time ago). Benefits nobody
(except maybe SCO) now. Nobody is really taking it seriously in the h/w space-
we'd be better off collectively pooling $$ and joining all of *BSD as an I2O 
working group member if we care about gettin the h/w folks onboard..."...