Welchman made a thorough study of the traffic, taking note of the radio frequencies and times of origin as well as the discriminants, call-signs and messages sent in parts. He marked the different networks with red, blue and green crayons and when message-breaking for research purposes went into production in the huts in 1940 these colours were retained: red for the main Luftwaffe cipher, which was used widely for operational and administrative purposes, green for the cipher used by the German army’s military districts, and blue for the Luftwaffe practice cipher. So far so good for Welchman’s organisational skills, but he, like Dilly, was an ideas man and he mistakenly thought that Dilly had deliberately ‘banished’ him to Elmer’s School away from the main cryptographic stream. ‘During my first week or two at Bletchley, I got the impression that he didn’t like me,’ Welchman said. ‘Very soon after my arrival I was turned out of the cottage and sent to Elmer’s School.’
Welchman recalled having noticed the formation of the occasional repeat letters in the indicators at the beginning of the message, which both the Poles and Dilly had picked up and exploited in different ways, leading to the present all-out punching operation in the Cottage. Independently, he had come up with the same idea for what Dilly called ‘females’ in the machine cycle and rushed over from the School to the Cottage to tell Dilly about it. Dilly was furious, he said, and reminded him that he had been told to study discriminants and call-signs, not methods of breaking the Enigma.
The trouble with reminiscences forty years on is that while Welchman gives a very accurate account of his thought processes and technical achievements, his memory fails him on incidentally related matters and he did not have the benefit of being able to check them with the now released official records. From personal experience of my own Bletchley Park memories, I am painfully aware how essential that is, considering that our memories had to remain buried for over thirty years.
The records show that Welchman was not ‘banished’ to the School because, as he thought, Dilly did not like him. Where Dilly was concerned, lack of communication was not a sign of dislike but merely of total absorption in a project to the exclusion of all else. Penelope Fitzgerald, Dilly’s niece, was told, very probably by Peter Twinn, who was aware of the situation, that when Dilly was told about Welchman’s perception, he said apologetically: ‘Hadn’t one said the right thing?’ The response was: ‘You haven’t said anything; you haven’t spoken to him at all for two weeks.’ Dilly was aware of how necessary and important the work of analysing the new wartime traffic would be and knew that Welchman had the right skills to do it; skills which would, thankfully, leave him to get on with the burden of all the tasks he had set himself. Nor was Welchman left without any explanation of what was going on in Dilly’s Enigma research. Kendrick was no lightweight and it was he who normally gave fundamental explanations about the Enigma machine to newcomers, although he did not himself make any inspirational contribution to codebreaking. Hugh Alexander, the head of Hut 8, later commented that any new suggestion in the hut had already been proposed by Kendrick, who had worked for Dilly and knew all the answers. A confidential note of Dilly’s in November 1939 says that Kendrick is ‘quite admirable’ and the ‘obvious second-in-command’ and it was a pity he was in Elmer’s School. Welchman, he said in the same note, was ‘doing well and is keen. I hope to get him back here to learn about the machines.’ Several documents from 1 November 1939 onwards show that, although Welchman was not then actually working in the Cottage, he was considered to be part of Dilly’s team as the documents are jointly signed by Knox, Twinn, Turing, Welchman and Jeffreys.
Dilly’s relationship with Alan Turing was rather different. Turing arrived in the Cottage with a full understanding of the history of breaking the Enigma, having worked on it himself in the summer and having had the post-Warsaw discussions with Dilly. As soon as Turing arrived in the Cottage he wanted to tackle naval, which Dilly had put aside in 1937 to await a ‘pinch’ of bigram tables; this suited Turing as he could work on his own and he disappeared into the stable loft, away from the constant bustle of making the perforated sheets in the rooms below. He did not want coffee breaks or social meals in the mansion so Claire Harding and Elizabeth Grainger rigged up a pulley to send up coffee and sandwiches to him in a basket. As with Dilly, his ideas sparked off in all directions, which makes Dilly’s confidential note on Turing a true case of the pot calling the kettle black:
Turing is very difficult to anchor down. He is very clever but quite irresponsible and throws out suggestions of all sorts of merit. I have just, but only just, enough authority and ability to keep his ideas in some sort of order and discipline. But he is very nice about it all.
Ever since Room 40 days, Dilly’s heart had been in naval intelligence and now that the U-boat war had started he was determined to do all he could to help Turing find a crib to solve the naval Enigma. His first opportunity came in November 1939, when he managed to be present at the interrogation of a recently captured naval prisoner of war, who might have had access to cipher material. Admiral Godfrey, the director of naval intelligence, thought it would be useful to have Dilly present at the interview. The information Dilly managed to extract from the sailor was indeed helpful. As there were no numbers on the Enigma keyboard, the German navy had hitherto used the letters QWERTZUIO for the figures 1 to 9, with Ys as separators, so that when a message was a continuation (FORT) of a previous one, the time of origin at the beginning of the message, if 13.35, would appear in the format FORTYQEETY, where FORT for continuation, is followed by a space, indicated by a Y, followed by Q for 1, E (twice) for 3 and T for 5 followed by one more space. Now Dilly was told that the Enigma operators had in future to spell out numbers in full, which made for a much longer crib.
Turing was working on ideas for mechanised cryptography as well as attempting to solve the naval Enigma messages. Although his work was previously theoretical, he was mechanically apt and had once made an electric multiplier and later designed a mechanism for answering a problem dealing with prime numbers, cutting the gears himself. As soon as Dilly had got back from Warsaw, he wanted to tell Turing about the Polish idea, which was new to him. At Pyry, he had seen what the Poles called their bomba (bomby in the plural); this was an electro-mechanical device, superior to a cyclometer, based on the Polish Enigma models, given its name because the idea came to Marian Rejewski and his colleagues while eating an ice-cream dessert called a bomba. The bombe, as it was called in French, was made by the AVA Telecommunications factory in Warsaw in November 1938 to cope with the changes in the German Enigma service messages. Drawings and plans of it had arrived at Victoria station on 16 August 1939 and, along with the Zygalski sheets and the model of the Enigma machine, were now in the Cottage for Turing to study.
Before leaving Warsaw Dilly had commented that the Polish methods all depended on the double encipherment of the indicators, ‘which may at any time be cancelled’, as it indeed was in May 1940. Dilly had discussed his ideas for electro-mechanical machines similar to the Polish bomby on his return from Warsaw, telling Turing that they should be thinking of developing a bombe method to check for standard beginnings of cipher texts and indicators. A few weeks after setting up in the Cottage, Dilly obtained permission from Alastair Denniston to acquire the former plum store in the stable-yard as ‘a small experimental workshop for trying out gadgets’. It was here that Turing’s ideas took real shape. Dilly was well pleased with how Turing, whom he called ‘my bombe-ish boy’, developed the British bombe, an electro-mechanical machine which ran through all the various possibilities of wheel choice, order, ring position and machine settings at high speed to test whether favoured positions by the codebreakers worked. This would greatly speed up the breaking of Enigma messages and the intelligence derived from them.
A meeting on 1 November 1939, attended by Dilly, Twinn, Jeffreys, Turing and Welchman, set out their requirements; namely punches, cyclometers and a British variant of the Polish bomby. Denniston’s deputy, Commander E
dward Travis, liaised with the British Tabulating Company at Letchworth about ordering the machinery. Dilly’s relationship with the bulldog-like Travis was hit-and-miss as both he and Denniston seem to have been paying unwelcome attention to the activities in the Cottage, which Dilly doubted they understood. The cyclometer was essential for testing female positions and consisted of two un-Steckered Enigmas wired together so that the output of each went straight into the input of the other. A Dilly memorandum stated:
What happened precisely was this. I found that the cyclometer was going very slowly and could not be speeded up to give more than four or five wheel orders. I told Jeffreys of my scheme and suggested that we should put on the best crew and get on with the scheme of check sheets. He demurred that it was against orders. I asked him what he thought of the scheme itself. He said it was the only sane thing to do. I threw about what little weight I have. I said I was his immediate superior; he was to keep the whole thing a secret; if AGD or Travis came in he was to conceal what we were doing as best he could. If God Almighty came in he was to report him to me.
Travis also took active steps for the provision of huts for the increased staff that would be needed when current German Enigma messages were broken. His report to Denniston on 19 November 1939 shows he already had the structure of the research and production departments well in hand. Travis was on good terms with Welchman, who now had two staff working for him on traffic analysis in Elmer’s School, allowing him to give much thought to how Hut 6, which he was destined to lead, could operate as the production line for the German army and Luftwaffe Enigma decodes. He estimated that it would be his red (Luftwaffe) traffic, which, owing to the increased number of intercepts, was likely to yield the first messages to be broken. As early as 29 October, Welchman had stipulated that the hut dealing with red would need a registration room, an intercept control room, a decoding room for the messages to be run off when the day’s key setting was known, and the sheet-stacking Netz room. A plan he drew up for Travis was so detailed that it even showed where the light plugs should be placed in the skirting board for the bulbs under the glass-topped tables for inspecting the perforated sheets. As yet Turing’s bombes were wishful thinking but the October plan showed where power plugs should be put in readiness. Dilly was preoccupied with other things at that stage and was happy to be left out of the production line discussions.
Thanks to Welchman’s painstaking traffic registration, Dilly had been presented with many more cryptographical opportunities. ‘Cillis’ were one of Dilly’s most important discoveries and would have far-reaching consequences. They resulted from a combination of two different mistakes made by some Enigma operators. While breaking pre-war traffic Dilly noticed that the operator often took the final position of the three wheels at the end of the preceding message as the setting for the next, thus saving himself the trouble of altering the wheels for the enciphering of the next message setting. This could be identified from the outside indicator in the message preamble. The first error the operator had to make for cillying to work was using a setting that could be guessed in the first message: the use of keyboard ‘slides’, where three letters next to each other on the keyboard were used – such as UJM or TGB – or ‘pronounceables’, the first three letters of a proper name, for example WALter, MARtha or CILli, in the first message; it was the discovery of an example of that last one in an early message that gave the name to the process. By subtracting from the outside indicator of the second message the number of letters in the preceding message it could be seen whether the first indicator had been cillied. The subtraction was done by sliding measuring strips, the number of units of length being the number of letters in the message, against a long strip consisting of the alphabet written out several times. Dilly in his playful way called these snakes. Other examples of this ‘Dillyese’ of the period include corsets, alligators, slugs, grass skirts, starfish and beetles.
Dennis Babbage arrived in the Cottage in December 1939 in time for the punch party and was present for the brilliant cilli discovery. He became an expert at cillying himself and found a new category where a lazy operator might even use the same three letters both for the indicator and the message setting, which was dubbed JABJAB. This all led on to the brilliant Ringstellung ‘tip’ by John Herivel in conjunction with cillis, where the lazy operator, having duly set up the clips on the wheels, when the key was changed at midnight shut the lid of the machine, allowing the letters to show in the windows, which he did not change when he came to send off his first message of the day. This became known as the ‘Herivel tip’. Hugh Foss had invented sheets which would show any clusters of giveaway indicators in network settings. It was a simple sheet of squared paper with alphabets at the top and left-hand side. A three-letter indicator WER would have R written in the square marking W at the top and E at the side. Testing for self-Stecker and reciprocal pairings when the wheel order was known, by using an un-Steckered machine, was always a ‘great thrill’, Babbage recalled, all part of the ‘magic’ atmosphere of Dilly’s early cilli handbreaks, the ‘miracle’ that became routine in Hut 6.
All the knowledge they had was based on pre-war information, however; it was not certain that the Germans had not changed the Enigma machine in readiness for the outbreak of war and there was in fact some indication that they had done so. Dilly carried on with the preparation of the Zygalski sheets, concealing from the exhausted staff that they might be useless for current traffic if the wheel wirings had changed. Fortunately, feeling so strongly about his promise to the Poles, he made one of his frequent threats to resign unless the final batch of Zygalski sheets was sent to Bruno.
The Cottage
7 January 1940
My dear Denniston,
As you remember on our Journey to Warsaw I promised to assist the Poles and the French in producing statistics. Actually we have produced these, or similar statistics, and in a third of the time that all three parties could have been expected to produce them, but I hold, if only on personal grounds, that these statistics (under copy) must be handed over at once.
They were delayed for two reasons only:
1. To devise a scheme of mechanical reproduction. This problem has been solved without using the Netz themselves.
2. In order to see whether the upper and under copies were needed at once. The answer here is ‘No’.
My personal feelings on the matter are so strong that unless they leave by Wednesday night I shall tender my resignation.
I do want to go to Paris but if you cannot secure another messenger I’m actually at the moment completely idle.
Yours ever
A. D. Knox
They were finally taken over by Turing on 17 January and given to Zygalski. Turing also met Marian Rejewski and told him of Dilly’s recent successes but was evidently anxious not to discuss the bombe. Rejewski, ignorant of how much progress he had made, recalled that they treated Turing as ‘a younger colleague who had specialised in mathematical logic and was just starting out on cryptology’. However, Rejewski was able to tell Turing why they had been unable to break wartime traffic with the Zygalski sheets and in his presence decoded a green message of October 28 with the new sheets Turing had brought over. Inadvertently, the Poles had given them some wrong information about the turnovers in the new wheels IV and V, which the Germans had introduced at the end of 1938. Dilly’s cillis depended on wheel turnover matches, which accounts for his inability to make his crucial method work on the 1939 traffic. Immediately on Turing’s return, the Cottage got to work on a green message for 25 October. Cillying had reduced the number of wheel orders for these messages to only three. This was the first wartime German Enigma message to be broken in this country and was received with shrieks of joy in the Cottage.
However, there was still the anxiety as to whether the Germans would make other changes in 1940 and it is best to let Babbage, who was there, describe the atmosphere in his own words:
We eagerly awaited the opportunity of finding th
e answer to the next great question. Had the Germans made a change in the machine at the New Year? While we awaited a suitable day, that is one with enough females for our purpose, several other 1939 keys were broken and we began to get evidence of the extent and nature of cillying. At last the favourable day arrived, and it had, besides the requisite number of females, several good cillies to cut down the wheel order. The sheets were laid, the stories tested and ‘Red’ of 6 January 1940 was out.
Up until now, a use for the Jeffreys sheets, as opposed to the Zygalski sheets, had not been found as they were designed to work with rodding, which was only possible with un-Steckered machines or on rare occasions where the Stecker were known and could be stripped off, as with the Enigma manual with its genuine crib. Dilly’s next observation detecting such a situation is a splendid example of his long experience in being prepared for anything, however unexpected, in the way of procedural errors which would give a way in. He had observed a simple meteorological telegram transmitted at midnight from German airfield radio stations in three letter groups, presumably taken from a codebook, which did not take him long to compile. Before long it appeared that someone had had the bright idea of adding a layer of cipher for increased security and Dilly noticed at once that this was an unusual reciprocal substitution, which suggested possible Stecker connections that the operator could have obtained from the daily Enigma settings.
Colonel Gwido Langer visited Dilly’s section in December 1939 to sort out arrangements for liaison with Bruno and the new discovery was soon shared with the Poles. As the weather messages were always transmitted at midnight, knowing the Stecker for a given day was appreciated as an early start to finding the Enigma settings. With the Stecker removed, Dilly’s rodding was possible and the Jeffreys sheets came into their own. If there was a crib guess of cipher and text involving rod couplings on the right-hand wheel, those sheets were superimposed and any hole alignment would indicate a possible combination and position of the first two wheels for which all these couplings were valid.
Dilly Page 13