Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!pasteur!ucbvax!BU-CS.BU.EDU!bzs
From: bzs@BU-CS.BU.EDU (Barry Shein)
Newsgroups: comp.protocols.tcp-ip
Subject: SLIP suggestion
Message-ID: <8804061437.AA14141@bu-cs.bu.edu>
Date: 6 Apr 88 14:37:51 GMT
References: <8804060515.AA14072@bu-cs.bu.edu>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 44


>Has anyone considered making SLIP work on an IBM system?  In particular,
>most IBM systems will only handle 7-bit ASCII.  When I last looked at
>SLIP it assumed an 8-bit channel.  IBM reads must be terminated by a
>carriage return also.  There might also be full/half duplex
>compatibility problems.
>
>Neil Erdwien
>Kansas State University

Gee, took a minute to figure out if this was referring to a PC or
what.

My guess is you're referring to a 370 architecture mainframe running a
line to a 3705 or equivalent. I've never done SLIP but I did write a
serial link of my own design once a few years ago between a 4.2 Unix
(probably 4.1 back then) system here and a 3705 which transferred jobs
for printing or batch execution.

Basically I found the only way to deal with the 3705 was to write a
layer which passed the packets as discrete ascii lines of text (I just
translated to hex ascii, prepending each line with a checksum and byte
count) and lock-stepped out the packets (send/ack/send/ack etc., just
like a card reader.)

The half-duplex nature of the interface presents some interesting
problems. Actually, it's more like quarter-duplex, it won't even
listen unless it wants to (you have to lock-step on a prompt from it,
usually a dot in column one.) If you talk before it's expecting you to
it throws away the characters and then demands a carraige return to
assure it you won't do that again (at which point you are free to
start the error cycle all over again.)

It's not ideal (hah!) but it can be done, once I got it working it
worked pretty much unnoticed, but there were a lot of heuristics in
there (did we just get a "!:" oops, it's angry! send a <CR> and back
off for a bit waiting for a "<CR><LF>.").

You also might want to see if there's a mode buried in there that was
used for downloading from a KSR33 paper tape which did flow control
(used ^S/^Q to express it's distaste for your data.)

Or get a network interface...

	-Barry Shein, Boston University