HDB UUCP Problem

[Bill Kennedy]

I will not repeat Sandy's or Roger's remarks about the ctsflow and
-ctsflow phenomenon.  In particular, however, they're complaining
about their stty getting unwound by uucico and throughput problems
as a result.  I emailed Sandy a suggestion and he confirmed similar
results.

My systems are a '386 AT clone with AT&T 386 UNIX Vr3.1 and a '286
AT clone with Microport V/AT 2.4.  Each is Trailbalzer equipped,
neither does a very good job with CTS flow control.  I did have one
of the TB's on a smart card that speaks CTS flow control but had
terrible performance at 1200bps.

I am routinely and reliably performing uucp transfers with the interface
locked at 19,200bps, auto-baud on the analog side and *NO* flow control.
I have a ghastly hack in /etc/profile for login users < 9600bps that does
set XON/XOFF handshaking, it gets reset with the DTR transition that
concludes the session.  I was very skeptical about this set up until one
uucp neighbor (a Trailblazer) complained about crummy transfer rates
and another (1200bps only) did too.

I believe that the phenomenon Sandy and Roger are complaining about is the
uucico setting raw mode, thus defeating ctsflow.  I have found, and Sandy
confirms, that the uucico's pace each other at any modem speed and that
throughput actually _increases_ when you remove all flow control.  That is,
of course, not true for a login user running vi at 1200bps :-) but the
XON/XOFF flow control handles that one.  I know you're talking about Xenix
here, and I wouldn't have brought it up had Sandy not said it worked.  My
observation is based on ~20hrs off hook a week with no complaints.
-- 
Bill Kennedy  usenet      {killer,att,cs.utexas.edu,sun!daver}!ssbn!bill
              internet    bill at ssbn.WLK.COM