mailing list archives
Re: NSF and the Birth of the Internet
From: David Farber <dave () farber net>
Date: Wed, 20 Aug 2008 08:17:54 -0400
Begin forwarded message:
From: "Edward Vielmetti" <edward.vielmetti () gmail com>
Date: August 19, 2008 9:03:45 PM EDT
To: dave () farber net
Subject: Re: [IP] Re: NSF and the Birth of the Internet
Dave, for IP if you wish:
One aspect of internet history perhaps underappreciated is how
fragmented campus computing networks were at one time, and
how the development of national networks were a spur to making
cross-campus communications easier.
The Internet I found when I first wandered into a housemate's
Usenet account in 1985 or so at Michigan was structured
so that email left campus three ways:
MTS mainframe mail via MIT-MULTICS.ARPA in Massachusetts
EECS Unix mail via ihnp4 in Illinois
Physics mail via SPAN
and thus cross-campus email from EECS to MTS required baroque
and completely unreliable addressing.
The on-campus postmaster duties that occupied most of my undergraduate
career involved systematically simplifying on-campus email and
making it more reliable, with the overriding goal to allow any
department head to send email to any dean in the morning and
have it received in time to schedule a lunch meeting.
There was a lot of work to do, in part because on-campus
email volumes only increased as more people got online,
and because expectations moved from same day service
to instant service. Here's some of the things we had to do.
* The MTS mail system had a front end gateway, a Vax running
BSD 4.2 connected to it over an X.29 line. I was the junior
postmaster there making sure that the mail queues kept
moving. Sendmail incorporated a bunch of changes to handle
queue managment over that time, since intelligent handling
of bounces, temporary network failures, and slow links made
a huge difference in throughput. That system eventually got
a direct TCP/IP connection which had a big impact.
* Name servers had to get set up, loaded with current information,
and debugged and bullet proofed to the point where the whole
campus could rely on them. Hans-Werner Braun was responsible
at the start of the umich.edu DNS setup, and I remember at
least a few late nights where patches went in more promptly
than you'd do now.
* Various and sundry Unix systems around campus had resolver
libraries patched in so that they could use the (then new) domain
name system. The then-current SunOS shipped with name service
through hosts.txt, so you had to fix each system that you wanted
to turn up.
* There was a considerable amount of UUCP going on around
the area, both to support Usenet newsfeeds and to connect to
sundry industrial affiliates (and hangers on). Maintaining the UUCP
network was part of the professional training of systems administrators,
making sure that they knew where to get their next job (at the
other end of the email link).
Noticably delayed on the Michigan campus was BITNET; that was
eventually turned up near the end of the mainframe era there (1990?)
after I had left campus.
Someone needs to also mention the Morris Worm - 20th anniversary
coming up soon! - which I witnessed first hand while doing
tail -f /var/log/mail.log
of the log files on the Vax from my Apollo Domain console late one night
and seeing weird error messages. Fortunately most of the campus
mail systems had enough patches applied that the precise infection
vector didn't work.
Edward.Vielmetti%UMICH-MTS.MAILNET () MIT-MULTICS ARPA
RSS Feed: https://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com
- Re: NSF and the Birth of the Internet, (continued)