Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!cs.utexas.edu!uunet!auspex!guy
From: guy@auspex.auspex.com (Guy Harris)
Newsgroups: comp.mail.uucp
Subject: Re: g protocol
Keywords: g uucico
Message-ID: <8053@auspex.auspex.com>
Date: 27 May 91 19:42:22 GMT
References: <2174@kgw2.XETRON.COM> <1991May25.115657.22742@skypod.guild.org>
Organization: Auspex Systems, Santa Clara
Lines: 18

>How many protocols are there?  I normally use g, but I notice that I also
>have an "e" protocol for error-free lines.  Would you use that if you
>are using error-correcting modems (like V.42 or MNP 4)?

I probably wouldn't.  The "e" protocol assumes not only that the *line*
is error-free, but that the path from the line to "uucico" is error-free
*and* that the entire path is flow-controlled.

This means that the path from the line to "uucico" mustn't drop
characters, and must use some kind of flow control to the other system
to make sure that doesn't happen; this means that it shouldn't pass
through a typical UNIX serial port driver, because those drivers drop
characters if they get too many characters (more than then can buffer),
and the only form of flow control that they'll do is to send a ^S or
maybe fiddle with some modem-control lines, and unless you can trust
that same form of flow control to the same extent that you can trust
TCP's flow control, it's not good enough for "e".  (The "e" and "t"
protocols are intended to run atop protocols as reliable as TCP.)