> Looks like it's off by something other than but near 2. :-( I'd suggest
> trying it at magma@600 and ttya@1200 or something like that. Then try
> magma@ nearby values. The advantage of doing it slower is that the baud
> rate generators have valid rates which are closer together, so you're more
> likely to be able to match divisors.
> 

I may have two different problems brewing:

- I went back to the "old" model card (the one with dip switches,
http://www.magma.com/sbus/21splc.gif) and made a fresh device (mknod
/dev/ttym0 c 100 0) and for some silly reason I'm able to use it now from
tip, but pppd refuses -- claims it's not 8bit clean. 

- The "new" model card (no dip switches, if you can pardon my cheesy grab 
http://peter.boku.net/images/21sp.bmp) included an addendum sheet that
denotes that it can go faster (460.8kb, up from 128kb) so they may have
gotten tricky and there is a flat-out mismatch.  This is further supported
by the fact that when I downloaded openbsd 2.5, it had the exact same
symptoms as netbsd 1.4.

So I finally cracked the book that came with the card and they have played
a trick so older OS' can use the higher speed interfaces.  These cards are
new to me but there is a 'setport' utility that comes with their drivers. 
Is there such a critter on netbsd?  Does the NetBSD driver "do the right
thing" and I'm I wandering too far afield?