>From: Thor Lancelot Simon <tls@cloud9.net>

>You might be able to pound the driver to do hardware flow control right,
>but I'd still be surprised to see a 3100 talk to a serial port at 9600 while
>doing anything else -- and unless the UART in your modem can handle 3% fuzz
>in the inbound data rate, you're not going to get anywhere at 19.2, though as
>I recall that goddamned DZ11-on-a-chip can sometimes, if you're lucky,
>recieve okay at 19.2 -- but don't count on it.

O tempore, o mores.

If one is *really* dedicated, there is a solution.  If an 11/750 can
keep up with a second unibus full of DZ-11s, a paltry R2000A should be
able to keep up with one DZ-11.  Use the 4bsd solution: pseudo-dma, in
hand-tuned assembler, jammed into the interrupt handler.  (Thank
goodness GCC has better support for this than PCC did.)  This doesn't
address bit-clocking-fuzz problems, though.

Of course this would require more person-hours than it's worth;
buying a cheap x86 should be a better solution. Or even
an 8-serial-ports-on-a-scsi-bus device, if you can get the interface
spec (they tend to be proprietary and non-disclosure) to write
a NetBSD driver.


Insert :) as appropriate. It's left as an exercise to the reader which
parts need them, and which don't.  And now back to our regularly
scheduled broadcast...