Tony Sale was also responsible for jump-starting the Colossus Rebuild project.

Colossus was used in the last War to break the German's Lorenz
cipher This encrypted a message written via teleprinter. Unlike the famous Enigma coding machine, teleprinters used not the standard 26-character alphabet and Morse but 32-symbol Baudot code. Each code output is of five channels - represented as no-hole or hole, 0 or 1, dot or cross.

Lorenz based his encryption system on Gilbert Vernam's 1918 design, whereby the text was enciphered using a mathematical method known as modulo-2 arithmetic. Basically this system modified each character of the text using another 'secret' letter. The receiver of the encoded message would turn the message back into plain text by repeating the arithmetic process using the same secret character.

Vernam's 'ideal world' scenario was that the random obscurantist characters should be completely random and pre-punched on to paper tape, used character by character in sync with the input message character

However, in war (Cold or hot), it's is often difficult, if not impossible, to make sure that same random character tapes were available for sender and receiver and that they were both set to the same start position.

Lorenz made a machine to generate pseudo-random character sequence - being a machine and, therefore, tied to some form of logic, no true randomness was possible.

"Unfortunately for the German Army it was more "pseudo" than random and that was how it was broken."

The Lorenz cipher, codenamed Fish, was first heard by British intelligence in 1940, the messages being known as "Tunny". Two years of painstaking brain sweat later revealed the entire logic of the Lorenz machine (they didn't get to see the actual gizmo until the end of the war).

Because the Vernam system depended on addition of characters, it was reasoned that if the operators made a mistake and used the same Lorenz machine starts for two messages (a depth), then by adding the two cipher texts together character by character, the obscuring character sequence would disappear.

The sequence of characters left each represented the addition of the two characters in the original German message texts. For two completely different messages, it is virtually impossible to assign the correct characters to each message. Just small sections at the start could be derived but not complete messages.

But, on 30 August 1941, a Colossal cock-up. A German operator had a long message of nearly 4,000 characters to be sent He correctly set up his Lorenz machine and sent a twelve letter indicator, using the German names, to the operator at the receiving end. This operator then set his Lorenz machine and asked the operator at the sending end to start sending his message.

After nearly 4,000 characters had been keyed in at the sending end, by hand, the operator at the receiving end sent back by radio the equivalent, in German, of "didn't get that — send it again".!

Both machines were returned to the same start position. Absolutely verboten but they did it. The operator at the sending end then began to key in the message again, by hand. If he had been an automaton and used exactly the same key strokes as the first time, all the eavesdroppers would have got would have been two identical copies of the cipher text. Input the same — machines generating the same obscuring characters — same cipher text.

Being human, however, and thoroughly disgusted at having to key it all again, the sending operator began to make differences in the second message compared to the first.

The message began with a known German phrase SPRUCHNUMMER — "message number" in English. The first time the operator keyed in S P R U C H N U M M E R. The second time he keyed in S P R U C H N R and then the rest of the message text.

Now NR means the same as NUMMER, so what difference did that make? It meant that immediately following the N the two texts were different. But the machines were generating the same obscuring sequence, therefore the cipher texts were different from that point on.

The interceptors realised the potential importance of these two messages because the twelve letter indicators were the same. At Bletchley Park, using the same additive technique to this pair as to previous Depths, they got get much further with working out the actual message texts.

Trying SPRUCHNUMMER at the start led immediately to the discovery that the second message was nearly identical to the first. Thus, the combined errors of having the machines back to the same start position and the text being re-keyed with just slight differences enabled complete recovery of both texts. The second one was about 500 characters shorter than the first where the German operator had been saving his fingers. This fact also allowed them to assign the correct message to its original cipher text.

Now the corresponding cipher and message texts could be added together, character by character, revealing for the first time a long stretch of the obscuring character sequence being generated by this German cipher machine!

Frank Morrell, of the Post Office Research Labs at Dollis Hill, produced a rack of uni-selectors and relays which emulated the logic. It was called "Tunny".The manual code breakers laboriously worked out the settings used for a particular message, which were plugged up on Tunny and the cipher text read in.

But, as it took in the order of a month to crack Tunnies, the results were more often than not, of interest only to historians.

In 1943, automation came to the party. A mathematician, Max Newman, believed it possible to automate some parts of the process for finding the settings used for each message. After some trial and error with paper tapes, ab Post Office engineer designed an electronic system and, 1500 valves later, Colossus was unwrapped at Bletchley Park as a Christmas present for SHAEF High Command.

Interceptions showed Germans had taken the deception of invasion at Pas de Calais, not Normandy; the bulk of the feared Panzer divisions remained in the north, in Belgium.