[at ARPA in 1974-1975], we were trying to set up
a satellite link with Britain, and to deal with the
Post Office Telecommunications, who oversaw
telephones in the UK back then, was just a
totally different experience from anything else.
They wanted us to buy insurance covering their
whole plant, in case our IMPs set fire to their
equipment, or something. It was really weird.
“Their worst fear was that somebody in
Europe would call up, through some kind of a
network, to a British telephone installation, and
get through it into the Atlantic link and get to the
United States, and somehow bypass the 15 cent
toll. ‘Christ,’ I said, ‘this is just a research and
development thing. If we can make it work, if it
turns out to be a great idea, then we can figure
out about rates’. We desperately needed to
extend the ARPANET link to Stuttgart, to some
American military base down there—and they
wouldn’t let us have the one little link.”
“The specification for the ARPANET was a
1,000-character message in half a second,” says
Crocker. “Not in two minutes, not tomorrow or
something like that. So it wasn’t like the US mail
system. This was a novel approach to the use of
the underlying communications line based upon packet switching
and, fortunately, ARPA had enough money and the latitude to do
this as an experiment without a lot of interference or legal issues.
That was a huge consciousness-raising event, in the sense of
once that project got going, it triggered big reactions from the
French, the Canadians, and the British. Everybody said ‘we have
to have a network too’. And then the race was on.”TIPPING THE BALANCE
The year 1971 also saw the arrival of Terminal Interface
Processors (TIPs) which allowed users to dial into the ARPANET,
accessing databases stored hundreds of miles away without
having to travel. Many TIPs were Honeywell 316 computers, with
up to 56KB of RAM, each of which could support up to 63 ASCII-
based terminals. We’ve only mentioned Honeywell machines
so far, but part of the challenge of building the ARPANET was
to transfer data without knowing the underlying architecture of
the machine it would be displayed on. This slowed things down,
as interface hardware and network software were developed
for the DEC-10, PDP 8 and 11, IBM 360, Multics (an early time-
sharing operating system for General Electric mainframes), the
Honeywell machines, and more.
In 1972, Kahn undertook the first public demonstration of
the ARPANET at the International Conference on Computer
Communications at the Hilton in Washington DC. The idea was
to install a packet switch and a TIP in the basement of the hotel
and invite conference attendees in to have a go. All the big names
were there—Crocker, Postel, Davies, and lots of others we’ve had
to leave out for the sake of brevity, many from overseas.
Postel had a demo cooked up that saw him retrieve a file from
a computer in Boston via a host at UCLA, and print it out there
in the hotel. A glitch saw the file sent, not to the printer, but to
a robotic turtle built by MIT that demonstrated how a machine
could follow a computer program and move. Unaccustomed to
the input designed for the printer, the turtle began jumping up
and down, twitching and jerking.
And that wasn’t the only thing to go wrong. One user, trying to
follow instructions but instead attempting to access a computer
that wasn’t operational, received a ‘Host Dead’ error and cried
out, “Oh my God! I’ve killed it!” Two other users logged on to aUtah machine and started a text chat, only to
discover they were sitting close to one another.
There were early chatbots too, set up to talk to
one another as if a doctor was talking to a patient.
Hundreds of people tried the new technology,
but the biggest failure came when Robert
Metcalfe—later to co-invent the Ethernet—was
giving some AT&T VPs a guided tour. The TIP
crashed and, despite being down for no more
than 20 seconds, this left an impression on the
telephone company suits. AT&T was providing
the data link for the demonstration but believed
its circuit-switching technology was superior to
packet-switching. “They made no point of hiding
their joy,” Metcalfe is quoted as saying in James
Pelkey’s A History of Computer Communications
1968-1988. “Because this confirmed for them that
circuit switching was better and more reliable
than packet switching, which was flaky and would
never work. I had been working on this for two or
three years and it really hacked me off.” Despite
AT&T’s skepticism, the demonstration was held
to be successful, and the conference saw the
founding of the International Network Working
Group, under the leadership of Vint Cerf.
It was the following year, in 1973, that the first illegal online
act was perpetrated. As a government-funded network, the use
of it for personal reasons was prohibited, but UCLA’s Professor
Kleinrock sent a request for the return of his electric razor, which
he had left behind after a meeting in the UK. It would only get
worse: in 1978, Gary Thuerk of DEC sent a mass-marketing email
to 400 recipients, in breach of the rules in place at the time. He
claimed it resulted in $13 million in sales.
ARPA was a research agency and so when ARPANET was
declared fully operational in the summer of 1975, the Defense
Communications Agency took control. At about the same
time, despite ARPANET having had passwords protected by a
polynomial hash algorithm since 1971, encryption devices were
added to the network to support classified communications, the
first ones manually keyed, with later iterations supporting the
automatic management of cryptographic keys.
ARPANET kept growing and evolving, however, standardizing
around TCP/IP in 1980 (the same year that USENET went public
at the University of North Carolina and Duke) thanks to a DoD
directive. UCL and the seismologists at NORSAR left the network
in 1982 but kept a connection to SATNET, the Atlantic Packet
Satellite Network that was linked to both ARPANET and the
PRNET packet radio network in 1977. All of these were funded by
ARPA, and many were built by BBN.
ARPANET shrank significantly in September 1984, after the
network was restructured to send unclassified military traffic
onto its own network, MILNET. This saw 68 of ARPANET’s
113 nodes move to the new network, though they could still be
accessed via a small number of gateways configured so that the
two networks could be separated if needed. Researchers used
ARPANET as an internet backbone, but its days were numbered.
In 1985, the National Science Foundation Network (NSFNET)
was established, providing access to five supercomputing centers
and the National Center for Atmospheric Research (NCAR) using
a 56kbit/s link (expanded to 1.5Mbit/s by July 1988 and 45 Mbit/s
in 1991) and using TCP/IP. It became active in 1986, connecting
Princeton University, Cornell University, San Diego, Pittsburgh,
and Urbana-Champaign in Illinois, all recently created
supercomputing hubs, and NCAR in Boulder, Colorado. This
became the main network backbone for government agenciesRobert Kahn in 2005, when
he received the Turing Award for
his contribution ‘of lasting and
major technical importance to
the computer field’.getting connected
44 MAXIMU MPC APR 2022
© SAN FRANCISCO CHRONICLE/HEARST NEWSPAPERS VIA GETTY IMAGES / CONTRIBUTOR