From: Jon Agar <AGAR@fs4.ma.man.ac.uk>
To: history-of-computing-uk@mailbase.ac.uk
Date: Wed, 31 Jul 1996 13:56:38 BST
Subject: Colossus faces rebirth -Forwarded

   Financial Times, July 27/28, 1996, pp. I, II.

   Colossus faces rebirth into a world of dispute

      There is controversy over whether a top-secret
      codebreaking device was really the world's first
      computer. Christian Tyler explains

In Block H at Bletchley Park, three grey-haired electronics engineers were
sorting through piles of old telephone switchboard parts. Beside them a loop
of teleprinter tape was being whizzed past photoelectric cells at the rate of
5,000 characters a second, driven by the motor from an old domestic vacuum
cleaner.

At right-angles to the tape-drive stood two parallel frames of grey metal
festooned with objects that today's microelectronic boffins would have
difficulty recognising - transformers, resistors. capacitors and relays.

The machine emitted clucks of sound, flickers of light, and a gentle heat that
permeated the hut. Every half second, an old-fashioned typewriter, mounted
chest-high on a stand beside it, would clatter out a line of numbers.

This contraption is a recreation of Colossus, the top- secret codebreaking
device which helped unscramble the radio traffic of the German high command
during the second world war. Thanks to its blinking valves and clicking
switches, Winston Churchill was able to read Hitler's orders to his generals
almost day by day for the last 18 months of the war.

The man behind the resurrection of Colossus is Tony Sale, a computer expert
and former MI5 operative who once worked for Peter Wright of "Spycatcher"
fame. Its reconstruction, now in its final weeks, is not merely an act of
homage to the mathematical supermen of Bletchley who hastened the Allied
victory over the Nazis. Neither is it just a triumph over the official secrecy
in which the machine was cloaked until a few years ago. It is a working
demonstration of Sale's contention that Colossus was the world's first
computer.

It is a controversial claim, especially for scientists in the US who earlier
this year celebrated the 50th anniversary of the "first computer in the
world". Who is right?

In 1943, when the prototype Colossus was built at the Post Office's research
laboratory in Dollis Hill, north-west London, "computer" meant a person, not a
thing. It would have been more readily applied to Bletchley's 7,000
mathematicians, linguists, servicemen, clerks and secretaries than to machines
devised for codebreaking.

Yet the idea of automating calculation was an old one. An early proponent was
the Englishman Charles Babbage (1791-1871)). But Babbage's "analytical engine"
was never realised in his lifetime due to a lack of engineering skills. In the
1890s, data processing arrived. Payroll, census and other statistics were fed
on punched cards into machines which selected and sorted them and tabulated
results. By the 1940s the punch card "ran the world" and International
Business Machines was in the driving seat, said Doron Swade, curator of
computing at the Science Museum in London.

Meanwhile the theory of a "universal" machine to manipulate symbols had been
developed by Alan Turing, the awkward mathematical genius from Cambridge
University, in a 1936 paper "On Computable Numbers". Wartime brought Turing
and a flock of fellow Cambridge mathematicians to Bletchley Park, a
"Tudor-bethan" mansion in 580 acres which had belonged to City stockbroker Sir
Herbert Leon. Chosen as the government's top secret communications base - now
called Government Communications Headquarters (GCHQ) - it was a short walk
from the junction of the old Oxford-to-Cambridge and London railway lines.

There, in Hut 8, Turing supervised the cracking of German naval signals which
used the Enigma code. Enigma was a three-wheel (later four-wheel) encryption
machine like a typewriter, a military specimen which had fallen into Polish
hands. The story of Enigma, deciphered with the help of electro-mechanical
"bombes" designed by Turing, is well known.

But the code for which Colossus was invented to crack was another matter, and
here Turing's influence was only indirect.

To understand Tony Sale's claim for Colossus as the first computer, it is
necessary to understand the job it was built to do. The code was generated by
a machine called the Lorenz, a 12-wheel affair with a total of 501 metal lugs
on the wheels' circumferences which could be turned "on" and "off" to create
huge combinations of patterns. The German operators would change the settings
of the wheels for each message and the patterns on the wheels at least once a
month. Because it was used for top-security messages, known to the Bletchley
people as "Fish", the Lorenz was, in theory, a great deal tougher to crack
than the Enigma.

Lorenz was as near to random as a machine could then get without being so
cumbersome as to be militarily useless. There was no question of working
systematically through the combinations. There were 10+19 possibilities which,
as Sale said, would give even a modern Cray computer a severe headache. But a
machine is still a machine, and humans have an amazing ability to detect
patterns in machine "randomness".

The Germans did not use Morse to send their encrypted messages but a pre-war
teleprinter alphabet called Baudot in which each letter is represented by a
combination of five on or off states. It is a binary system which can be
transmitted as positive and negative impulses, written as dots and crosses or
0s and 1s, or transcribed as holes and spaces on a punch tape.

Once the Lorenz machine had been set up it would add a "random" letter to each
letter of the message text. This addition used a logical convention where
0+1=1, 1+0=1, 0+0=0 and 1+1=0 (repeat 1+1=0). The resulting garble would be
transmitted by landline, undetected, or by radio, when it could be
intercepted. By adding back the same string of nonsense letters on his own
Lorenz machine, the receiving operator revealed the message.

The first intercept of "Fish" was made in 1941, of German army messages
between Vienna and Athens. By July of that year the secret traffic was being
laboriously deciphered by hand, at the rate of one message every six to eight
weeks.

But on August 30 the cryptanalysts had a stroke of luck. A long message of
3,976 characters, which Tony Sale calls "probably the most important message
ever sent by the Germans", was sent twice with the same wheel setting. By
"adding" them together the codebreakers were able to infer the hidden message.
By January the following year the mathematician Bill Tutte had cracked the
whole structure the logic - of the Lorenz machine and its 12 multi-tagged
wheels.

Professor Max Newman then decided to automate the process. The patterns on the
wheels could only be worked out at this stage by brainpower. But the wheel
*settings* were revealed by comparing two, one containing an intercepted
encyphered messages, the other the patterns guessed by the codebreakers. The
tapes were run on a 24-valve device called, with typical Bletchley humour, the
"Heath Robinson", after the cartoonist who specialised in drawings of
fantastical gadgets. Once the wheel settings were broken, the message was run
through another machine for decipherment. This was the "Tunny". Though it
looked nothing like the Lorenz - more like an old-fashioned telephone
switchboard - it mirrored the logic of that encryption device.

It was difficult to keep two punched tapes running synchronously at up to
30mph without stretching or breaking them. A Post Office engineer, Tommy
Flowers, now in his 80s, had a brainwave. Why not do away with one of the
tapes by generating the patterns using an electronic sequence *inside* the
machine?

The first Colossus, a monster for those days, boasted 1,500 valves. It was put
together at Dollis Hill and carted up to Bletchley Park in December, 1943. By
the following February it was working, in time to inform Churchill that his
elaborate pre-invasion deception efforts were having some effect on Hitler.
Colossus was doing in two hours what it took humans up to eight weeks to do. A
2,500-valve Mark II was ready in time for D-Day, when, thanks to Resistance
sabotage and aerial bombardment of landlines which forced the Germans to send
their messages by radio, about 300 high-level "Fish" orders were being
intercepted and deciphered each month.

So sophisticated was the machine intelligence at Bletchley Park that the very
existence of Colossus was not revealed until 1970, according to Tony Sale.
After the war the government ordered 10 Colossi to be broken up - some say as
part of an intelligence deal with the Americans. One may have survived at
GCHQ's new home in Cheltenham until about 1960.

Gripped by a desire to assert the claims of Colossus, the former MI5 man asked
GCHQ to reinstate his security clearance so he could work on the project. The
parts could be found in any British telephone exchange up to the 1970s. Yet it
took until 1992 to get all the electronics declassified. Only last November
was Sale allowed to demonstrate the machine's ability to break the Lorenz
wheel settings. Even today members of the public are forbidden to operate
Colossus: some of its codebreaking algorithms are still, it seems, a secret.

Was Colossus really a computer? It had no stored program and was set up before
each run by means of plugs and switches on a board. It had virtually no
memory. But Sale claims it was a very early example of parallel processing:
separate logic calculations were being carried out on each of the five
longitudinal tracks of the tape.

Sale describes the machine as a large electronic valve programmable logic
calculator. "No lay person would argue that it is not a computer," he says.
Furthermore, because it wasted no time retrieving a stored program it was
almost as fast as a high-speed modern computer whose nominal work-rate is
1,000 times quicker. A simulation of Colossus which Sale ran on a
top-of-the-range Pentium PC took twice as long as the real thing.

In April 1944, while the first Colossus was rattling away at Bletchley Park,
the Americans were designing at the engineering school of the University of
Pennsylvania a machine to work out ballistic tables for artillery shells. This
was Eniac, the Electronic Numerical Integrator and Calculator, containing no
fewer than 19,000 valves, a number made necessary by the fact that Eniac did
decimal arithmetic, not binary logic like Colossus. (Neither was a "universal
machine" in the sense envisaged by Turing but both, as Donald Michie showed in
England and John von Neumann in the US, could be adapted to perform a variety
of operations.)

Eniac was too late for the war and indeed was out of date before it was built.
In February this year the US vice-president Al Gore, attending a ceremony to
mark the 50th anniversary of Eniac's first run, declared it to be "the first
computer in the world".

Professor Maurice Wilkes, who oversaw the construction of an Eniac-type
machine called Edsac (Electronic Delay Storage Automatic Computer) in
Cambridge in 1948, says today that the Eniac was the first large-scale
electronic computer. "And it was a real computer in the sense that it was an
arithmetical machine. It just wasn't a stored-program computer."

That concept, embodied in Turing's 1936 paper, was about to be made explicit
in a report produced before VE Day by the Eniac team under the other "father
of the computer", the Hungarian-born John von Neumann. He named this machine
Edvac (Electronic Discrete Variable Computer). "This was the first formal
description of a stored program computer," says Doron Swade of the Science
Museum. "It had all the ingredients - input, output processing and memory."
But Edvac never saw the light of day under its own name.

Meanwhile Turing was at the National Physical Laboratory struggling with red
tape to get a machine built to his own design. As Alan Hodges described in his
1983 biography, The Enigma of Intelligence, Turing resigned from the project
in 1947, the year the transistor was invented. The complete ACE was unveiled
in 1957 as integrated circuits were appearing on the horizon.

The first demonstration of a computer in the modern sense of the word occurred
at the University of Manchester England, on June 21, 1948. Most historians
agree that "Baby", a demonstration machine built by a young postgraduate
called Tom Kilburn under the late Prof F.C. Williams, was the genuine article
- the first working stored-program electronic digital computer. A party is
planned for Baby's 50th anniversary, too.

The rivalry is not unlike the search for the first man: a great deal depends
on your definition. "I distrust any and all claims to be 'first'," says Dr
Michael Williams, an historian of computing at the University of Calgary in
Canada. "If you add enough adjectives to something it will always be first."

Computer archaeology is a strange business. The capacity of Colossus could now
be stored on a single chip; today's computer scientists would not recognise a
thermionic valve if it winked at them. Yet the youngest can still shake hands
with the oldest, men from the computer Stone Age like Tony Sale and his
friends tending the flame in Block H.

Two photos+ Tony Sale with his recreation of Colossus, the device which helped
unscramble radio traffic of the German high command during the second world
war. Lorenz, the German code-generating machine, was first intercepted by the
British in 1941.

End+