Return-Path: <Jonathan_Hardwick@gs69.sp.cs.cmu.edu>
Received: from gs69.sp.cs.cmu.edu by K.GP.CS.CMU.EDU id aa10984;
          8 Aug 92 21:01:14 EDT
Received: from gs69.sp.cs.cmu.edu by GS69.SP.CS.CMU.EDU id aa11835;
          8 Aug 92 21:00:45 EDT
To: paul_parker@CS.CMU.EDU, terri_watson@CS.CMU.EDU, bovik@CS.CMU.EDU
Subject: More SLIP Experiences
Date: Sat, 08 Aug 92 21:00:35 -0400
Message-ID: <11833.713322035@GS69.SP.CS.CMU.EDU>
From: Jonathan_Hardwick@GS69.SP.CS.CMU.EDU

More random bits of information:

Registering In Andrew-land
==========================
If/when you get a permanent IP address from data comms (dc0m+@andrew)
they also assign you a name for your machine, which is of the form:
	surname.SLIP.ANDREW.CMU.EDU
It takes extra effort to make them change this into the name you *want*
for your machine.

CMUCS SLIP Servers
==================
Help@cs has started offering SLIP service over the new Cisco boxes.
This has the advantage of taking data-comms out of the loop, letting
your home machine have a unique IP address with the name you want, and
(I think) shortening the path the bits have to take.  The downside is
that it doesn't *currently* offer Compressed SLIP, whereas the Andrew-
land service does.  CSLIP should make it to the Ciscos in a couple of
months.

Compressed SLIP
===============
Van Jacobsen came up with some neat tricks to trim a TCP/IP packet
header down from 40 bytes to 5 bytes, and called the result "Compressed
SLIP".  When you consider that each character typed at a telnet session
goes in its own TCP/IP packet, this has obvious performance advantages
for interactive use.  It makes gnu-emacs usable at my typing speed in
full X-windows mode; the CMU implementation sends some obscene number
(8?) of X events for every character typed in gnu-emacs, which
translates to a whole bunch of TCP/IP headers and not much data.

Anyway, a SunOS implementation can be found in the file cslipbeta.tar.Z
on any archive site -- you need to have the normal SLIP code as well.
Once again, versions exist for 3.x and 4.0.x systems.

Fine Tuning SunOS
=================
Van Jacobsen's comments in the CSLIP README file confirmed what I had
heard from other sources : the serial handler in SunOS 4.x sucks
compared to that in SunOS 3.5.  If you have a choice, stick with SunOS
3.5.  If you don't, check out the official Sun patch file YAPT5.5c (in
the file slipware.tar.Z, again findable via archie).  This fixes a
number of bugs and upgrade the subjective performance from "bearable" to
"reasonable".  Note : the patches were meant to be a part of SunOS
4.0.3, but never made it into the copy I inherited.  Double-check if
you're not sure...

Jonathan H.
