>> >  I don't recomend 57600 for serious data
>> >transfer in. 38400 seems ok.
>>
>> I wouldn't even use 38400...the transfers screw up at or around 9600, even.
>> 2400 baud seems very stable, as far as i can tell. slightly more errors
>> when i get to 9600, but over 19200 pushes the limit on the number of errors
>> you can stand :-\
>> running ppp over the device is something else, but the getty driver can't
>> handle it. (from my experience, anyway, like trying to use slirp or zmodem
>> file xfers)
>
>Strange. I must admit that I use ppp whenever I care about transfering
>data. I haven't used zmodem for about 18 months.

I'd have to agree that this is indeed strange.  I once had my NetBSD/mac 
connected to my laptop via the serial port using a null modem cable and 
ran the port at 57600.  This was with a 1.2D vintage kernel (not that 
out-of-date) using PPP (but I've also done the same with SLIP connections 
from my MacPlus).  In addition to doing many megabytes of FTP file 
transfers in both directions, I had the filesystems NFS mounted on the 
laptop where I was doing Kernel builds.  In about three days of 
continuous running I think I saw fewer than 1/2 dozen SILO overflows on 
the laptop and none on the Mac side. The Mac in question was a 68030 
running at 33Mhz which is a bit faster than the SE/30 (a Performa-550 by 
the way), but I can't see that this would force the SE/30 all the way 
down to 2400 baud for reliable connections.  Maybe there's some 
handshaking that's missing in the SE/30 setup where this was observed?

-bob