joda@pdc.kth.se (Johan Danielsson) writes:
> cgd@netbsd.org (Chris G. Demetriou) writes:
> > This is not right.
> 
> I agree, but the puc-style `here we have a long list of cards that we
> know about' isn't very pretty either.

No, it isn't, hence my desire for a pointer to a standard that says
what can really be expected.

I'd really much rather say "communications/serial means foo" than do
the table that's there now.  But i've seen at least one board which
said "communications serial" but didn't have devices at the normal
places.  (I think it was one of the dolphin boards; they were kinda
screwy.)


> Guessing that a random serial
> card is based on something that is 16x50 compatible is probably not
> too far from the truth (and this is what puc does too).

Well, what it does is match but then print out a note about how to get
the card supported if it's likely to be supportable.  (unfortunately,
for e.g. losemodems, this seems to cause confusion.)

I'd rather do that than, say, assume the wrong thing and produce an
unusable system.  (that would keep people from installing, for
instance.  8-)


> A negative
> match might be more useful (cards that we know to be non-functional).

yes, but in order to do that I think i'd want to see some positive
statement that indicated that a sane set of criteria were expected.


I really, really, really dislike the way 'puc' works.  I _hate_ having
to waste space in the kernel to support an array of devices when the
user's likely to only have one or two of them to begin with, and to
have to waste more and more space over time as more and more matching
devices are found...


How many devices have you tried your 'com' pci attachment on?


cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.