In message <199605031642.LAA17855@ncremp.ag.iastate.edu>, flipk@ncremp.ag.iasta
te.edu writes:
>>	That may be.  However, it's also possible that 38.4 or 56K is
>>just too fast for a 386/40 all by itself.  I'd also like to run two
>>(or more) modems in the not-too-distant future.  I'm relatively
>>resigned to the fact that there is a limit to 'fast' serial chips.
>
>	[Note: this is not a flame or insult.]
>
>	This is nonsense. Remember the old TRS-80 Model III? Remember
>the old Moterola SWTPC (6800-based [NOT 68k!])? I had these two machines
>talking to each other at 19.2k using interrupt-based drivers with no
>problems. (The 6800 ran at 814kHz; yes, that's less than a megahertz.)
>
>	Of course, they weren't doing much of anything else at the time,
>and whenever the TRS-80 wanted to access the disk, it had to XOFF (since
>the disk interface on the Model III is byte-polled [ICK!]), but they were
>otherwise able to keep up.
>
>	And, I used to run PPP across a room to my roommate's 486sx/25 from
>a 386sx/25 running Linux, and easily pushed 38.4k. (No FIFO's on either end.)
>As soon as I switched to NetBSD, silo overflows crashed the link regularly.
>(I know, really cheap network, you get what you pay for, but college 
>students do that sort of thing.)
>
>	But I just can't imagine that 38.4k is "too fast" for a 386/40..

  I completely agree with Phil.  Even with an OS like NetBSD
(multi-user) I cannot believe that a little serial communication is too
much for the machine.  I fail to see why NetBSD has this problem with
serial communication when other OSes do not appear to.  I never remember
Linux complaining about SILO anything.

  Thank you.

---
Kent Vander Velden
graphix@iastate.edu