[net.unix-wizards] UUCP and flow control

eric@milo.UUCP (Eric Bergan) (11/21/84)

> UUCP and Flow Control
> 
> 	We have this problem as we run an Ungermann/Bass
> 	broadband system around campus which is not transparent.
> 	(we do a lot of TCP/IP so uucp is not critical to us)
> 	I thought about this problem and the fix I proposed
> 	(but never implemented) was to teach uucp to talk to
> 	a psuedo terminal with the back-end process on the other
> 	side doing whatever escaping is needed for the stream.
> 	A socket could accomplish the same thing (or of course
> 	yet another ioctl().) Obviously the UUCP at the other
> 	end would have to do the same thing but I believe this
> 	software could be incorporated exactly the same as
> 	one incorporates a new modem into uucico (actually simpler
> 	as once it opens the psuedo device the other process
> 	could pick up a lot of the work.) Any thoughts?

	I have also run into this problem with UUCP on broadband networks
(Ungermann/Bass and Sytek, so far). I got around the problem by implementing
a new protocol under UUCP which is basically the 'g' protocol, only it
splits up the data bytes so that it avoids the special characters (^S, ^Q,
and the first character of the "attention sequence"). Another approach,
which I have not investigated, is to have uucp disable the
flow control on the box and the attention sequence, make the connection,
use the full eight bit data path, and then rely on dropping DTR to break
the circuit and return the network to "command" mode. This should work,
but I haven't tried it.

-- 
					eric
					...!seismo!umcp-cs!aplvax!milo!eric

sasaki@harvard.ARPA (Marty Sasaki) (11/22/84)

The Sytek network has real problems with flow control. We have things
set up so that XON/XOFF flow control is the default. We have tried
turning flow control off on the line, and then turning it back on.
It turns out that you can't reliably do this, so the state of the
line is unknown. The solution of checking the line status sometimes
doesn't work either. The problem became so difficult, that we abandoned
using the Sytek for UUCP.

-- 
			Marty Sasaki
			Havard University Science Center
			sasaki@harvard.{arpa,uucp}