The John Vincent Atanasoff who worked at the Naval Ordnance Laboratory for the next seven years, through the war and then on several projects afterward, was the same man who had made his self-confident, energetic, innovative, and sometimes abrasive way through school, college, graduate school, and a successful teaching career. He worked unceasingly and impressed everyone who knew him, but he did not always fit smoothly into the navy’s way of doing things, nor was he always happy at the way projects for the navy started and stopped according to what seemed to him to be whims on the part of the admirals in charge.
Lura and the children stayed behind in the house on Woodland Street in Ames—Atanasoff commuted throughout the war, a thirty-six-hour train trip each way on top of a weekly work schedule that could run as much as sixty hours. At first, he was put to work on developing mines and depth charges that would operate acoustically rather than magnetically. At the beginning of the war, especially in the Atlantic, mines were used that detected ships by sensing magnetic waves. When they then exploded, the shock wave of the explosion would damage the hull of the passing ship. These mines could be cleared or disarmed by minesweepers dragging electrical cables through an area and passing a large pulse of electric current through the cables, a method called the Double-L Sweep. The navy wanted a mine that could be triggered by the sound of a nearby propeller or the clanking of the metal plates of a ship’s hull.
Atanasoff had never specialized in acoustics, but as always, he mastered the material with a few months of intensive reading, and soon he was in charge of the entire acoustics division at the NOL, supervising about a hundred men. He directed projects on the acoustical properties of explosives, acoustical detection, acoustical location, and numerous other topics. He was so inventive that he was later cited for, among other things, “his unusual imagination and exceptional mechanical ingenuity, his enthusiasm and indefatigable energy and zeal.” Part of his citation might have easily described the invention of the computer: he “has succeeded in conceiving of the solution to an urgent military problem which had been considered insoluble. Having conceived of the answer to the problem, he saw it through design, production, and test to the final timely adoption despite almost insurmountable obstacles.” Some of the men Atanasoff hired he knew from Ames—almost all research not related to the war effort had been shut down by mid-1942, so there was a large talent pool to draw from. In the midst of all of this, he still had time for hobbies, one of which was a comparative study of alphabets. On his visits to Ames, he tried to keep track of the progress of the patent application for the ABC, but his efforts proved frustrating—he was never able to find out exactly what the college was doing about the patent or to persuade them to move more quickly.
Atanasoff’s desk was in the Naval Gun Factory—it is an index of his powers of concentration that he got so much work done in the midst of a constant din. Atanasoff held a very high security clearance, so one day in the late spring of 1943, he was surprised to look up and see John Mauchly standing in front of him. Mauchly sat down and lit a cigarette, and, as far as Atanasoff was concerned, they had a pleasant conversation in which they first discussed what they had been doing since they’d last met almost two years before.
Mauchly had a lot to tell. He had become involved in a project at the Moore School, calculating trajectories for aiming large pieces of artillery. The proper aiming of a cannon had to take into account all sorts of factors: the elevation of the cannon and the elevation of the target, wind speed, wind direction, air temperature, humidity, and numerous others things. The variables were organized into firing tables, which were calculated by women employed by the Moore School working on Monroe calculators, but, as with Atanasoff’s calculations for the dielectric constant of helium, the calculations were tedious and time-consuming—the army was inventing and producing weapons faster than they could be put to use. At the Aberdeen Proving Ground, the Bush Differential Analyzer—the analog machine based on Babbage’s Differential Analyzer and invented by Vannevar Bush, the man who was now the head of National Defense Research Committee—was making some headway on the necessary calculations, but the Bush Analyzer could only solve differential equations with up to eighteen variables. Mauchly was aware of the army’s problem and now told Atanasoff about his new colleague, Eckert. He explained that the two of them were attempting to devise a machine that the army could use to make the necessary firing-table calculations.
Mauchly had submitted two proposals. The first, seven pages entitled “The Use of High-Speed Vacuum Tubes for Calculation,” was submitted in August 1941 and described “an electronic device operating solely on the principle of counting.” He suggested that it would do the same jobs as analog machines, but do them more quickly. The army authorities in charge of research at the University of Pennsylvania apparently did not understand what he was getting at and also did not consider him a serious contender for research funding—one man, Carl Chambers, is quoted by Scott McCartney as saying, “None of us had much confidence in Mauchly at that time”—a sentiment Atanasoff would have agreed with.
The pivotal figure in Mauchly’s career was a twenty-eight-year-old lieutenant, Herman Goldstine, who happened to have a Phi Beta Kappa BA in mathematics and a PhD in ballistics from the University of Chicago. Before being drafted into the army, he had taught at the University of Michigan. Once he was drafted, a former professor found him a position at the Aberdeen Proving Ground. Goldstine was put in charge of the firing tables. When he took over, each table took a month to produce. Goldstine’s first thought was to hire more women to do the computations, but when his wife, Adele, also a mathematician, set out to find more female math students (female math students could do the calculations and were not as essential to actual combat operations), she could find only a few. The Bush Analyzer was too slow and hard to maintain in working order. It was Goldstine who heard of Mauchly and his idea, and Goldstine who found Mauchly and asked him about it.
But neither Mauchly nor John Grist Brainerd, the Moore School’s liaison with the army, could find a copy of Mauchly’s seven-page proposal, now eight months old. At Goldstine’s behest, Mauchly, Brainerd, and Brainerd’s secretary put together as good a new proposal as they could come up with and took it to Aberdeen.
A major Allied setback that was not understood until after the war was the fact that the Germans also managed to crack English codes, specifically the code that routed convoys, Naval Cipher No. 3. Even though they did not have the benefit of a machine like the Bombe to do so in real time, they could often figure out the “size, destinations, and departure times,” according to Andrew Roberts, but “instead of recognizing the danger, the Admiralty put the U-boats’ remarkable success in intercepting convoys down to the advanced hydrophone equipment they used … Naval Cipher Code No. 3 was not replaced with No. 5, which the Germans never cracked, until June 1943.” The spring of 1943 saw the sinking, between March 16 and March 20, of twenty-seven Allied ships on their way from New York to Liverpool; 360 seamen died in the battle. Captain H. Bonatz, of the Beobachtungsdienst, a German naval code-breaking organization, later recalled, “The Admiral at Halifax, Nova Scotia, was a big help to us. He sent out a Daily Situation Report which reached us every evening, and it always began ‘Addressees, Situation, Date.’ ” The rote repetition of the first words of the communication enabled the Germans to break the English codes every day in the same way that the repeated three-letter signal had helped the Enigma decoders. At Bletchley Park, the decoders could tell by what they were decoding that the Germans had access to Allied coded information. But code breaking in Germany was fragmented among various services and commands—there was never a well-funded center for deciphering Allied messages like Bletchley Park.
Turing, at this time, was in the United States. He spent a while at Bell Labs, working on a method for enciphering speech, where he discussed his paper “On Computable Numbers” with Claude Shannon. Shannon, himself a graduate of MIT, had written his master’s thesis in 1937 on using relay switches to solve Boolean algebra problems. He also had the insight, like Atanasoff, that the binary arithmetic that relay switches represented would simplify information systems. His master’s thesis, written when he was twenty-one and published when he was twenty-two, is considered to be one of the most important, if not the most important, master’s thesis of the twentieth century. Shannon had studied neurology, too. According to Hodges, when Turing and Shannon shared their ideas about “thinking machines” in March 1943, “they found their outlook to be the same: there was nothing sacred about the brain, and if a machine could do as well as the brain, then it would be thinking—although neither proposed any particular way in which this might be achieved.” At the end of March, Turing returned to England on The Empress of Scotland.
It was in this context that Mauchly submitted his second proposal to Goldstine and Goldstine sought authorization from the army to fund the project—conditions seemed dire and the army was desperate enough to grant $61,700 (the equivalent of $750,000 in 2010 dollars) to Mauchly and Eckert.
According to McCartney, Mauchly and Eckert discussed their ideas casually—sitting around the Moore School and spending time drawing on napkins in a restaurant nearby. “A machine could be designed to do nothing but count the pulses of electrons, with the pulses representing numbers, and to crunch numbers in different ways to solve different problems. Instead of moving gears and wheels in a conventional calculating machine, Mauchly thought he could build a machine with no moving parts: only the electrons would course through the machine.” Eckert agreed with and was inspired by the idea of electronic calculation—he had already devised a method of calculating smokestack emissions that sent a beam of light through a cloud of emissions. The amount of light that got through was then measured, giving a reading on the density of the emissions. There is no evidence that Mauchly and Eckert kept a record of their deliberations or that they elaborated on the theory behind the ideas that they passed back and forth between their meeting in June 1941 and the submission of the first proposal in August 1942.
It was with his authorization in his possession that Mauchly came to visit Atanasoff at the gun factory in April 1943, but he said nothing about it. After chatting amiably for a while, he did ask Atanasoff a few questions about the ABC and about Atanasoff’s computer design ideas. Atanasoff, still underestimating Mauchly in several ways, was as forthcoming as he had been before. He felt, after all, that he and Mauchly were friends and that they were on good terms. He also had few opportunities to discuss his passion for electronic calculation. It was only later, after thinking about their meeting, that Atanasoff wondered how Mauchly had gotten security clearance to visit him—to just show up. Though he asked around, he never got an answer to this question more satisfactory than the vague supposition that possibly Mauchly had connections, since his father was a Washington, D.C., scientific eminence.
After the first visit, Mauchly stopped by off and on, always chatting in a friendly way about personal matters before asking a few specific computer questions. At one meeting, he asked about the progress Iowa State was making toward patenting the ABC, but Atanasoff couldn’t answer that question with any certainty—he was working so hard at the NOL that he had neither the time nor the energy to keep after the college. Nor could Atanasoff say that he had kept on top of recent developments in computing—he was simply too busy. It seems clear from these conversations that Mauchly was using his access to Atanasoff both to probe him and to gauge whether the computer he was developing with Eckert might turn out to be profitable. The visits went on for three years.
Frugality was never a feature of ENIAC, which began to take shape in a large unused room at the Moore School in July 1943. At first the engineering team numbered twelve—Goldstine, Eckert, and Mauchly oversaw the general design (with Eckert in charge). Other members of the team were put in charge of individual components and, since the army was in desperate need of the firing tables, the Moore School team worked with seven-day-a-week dedication. Eckert’s most controversial decision was to use vacuum tubes—at first five thousand, a number that grew to eighteen thousand (in part because the army, in its desperation, pushed Goldstine to expand the capacity of the machine). Such a number was unheard of, not only because vacuum tubes themselves were considered unreliable, but also because wiring so many together would amplify the malfunction of any single one. But Eckert was determined to use the tubes and decided to make them less prone to burning out by obtaining only the best tubes and then operating them at a much lower voltage than recommended, as well as never turning the machine completely off—the current could be reduced to a trickle to keep the tubes warm and to guard against the potential danger of thermal shock.
Eckert was dedicated to testing every part. According to Scott McCartney, in order to choose his wiring, “Eckert acquired some mice in cages and starved them for a few days. Then he put different kinds of wire in their cages to see which kind they enjoyed eating. The least appetizing brand was used in ENIAC.”
Eckert and Mauchly also decided to use a decimal counting system, sort of an electronic version of the Monroe calculator—if the number 345,679 was entered into the calculator, the counter in the ones column would flash nine times, the counter in the tens column seven times, the counter in the hundred column six times, and so on. But the tubes were much faster, of course, than a person tapping a calculator—a number would register in two millionths of a second. The advantages of speed were balanced by the dangers of unreliability, and so the machine, which was huge, had to have repairability built into it—it was so important that every tube be accessible in case it burned out that the machine was designed in discrete units with doors that opened into the mesh of wiring and tubes, and it took so much power to run the machine that Eckert had to include safety switches on every door to prevent electrocution. In addition, because the machine was decimally based, it could only add and subtract, not multiply, but Eckert’s idea was that it would be so fast that a binary number system would not improve overall performance and would require an extra piece of input- output hardware.
Like Zuse, Goldstine found help where he could—moonlighting telephone workers assisted with the wiring, Bell Labs supplied telephone parts and help with those parts, IBM designed a card reader for input and output. Goldstine, Mauchly, and Eckert seemed to work together quite well—Mauchly came up with the ideas but was considered by the others to be easily distracted. Eckert followed through, realizing the ideas in the machine and making sure that his designs were properly executed—he was noted for his perfectionism (and appreciated, in light of the expense and the danger of what he was putting together). Goldstine found the money, organized the personnel, and was the liaison with the army. He got along well with Eckert, but not well with Mauchly, who seemed like “a space case” to him. Eckert, it was clear to everyone, depended on Mauchly, but no one knew exactly why, even Mauchly. Since Mauchly was teaching at the same time, he wasn’t present at the building site as much as the others were. In 1944, when his teaching load was cut back so that he could work full time on ENIAC, his salary was cut from $5,800 to $3,900, leading to even more anxiety on his part. But he still felt that the project was his because he had originated it. It was at this point that he applied for a part-time job at the NOL, in the statistics department, and used Atanasoff as a reference. Atanasoff later said that he gave Mauchly a good recommendation more out of friendship than out of faith in his talents or expertise, but Mauchly got hired.
Mauchly mentioned his machine the first time in early 1944—according to Tammara Burton, “He looked Atanasoff in the eye and told him that he was building a new computer. The new computer, Mauchly claimed, isn’t ‘anything like your machine’; but is ‘better than your machine.’ ” When Atanasoff had asked about the new computer, Mauchly put him off, saying that it was top secret. Though Atanasoff’s security clearance was higher than Mauchly’s, Atanasoff knew he would not get anywhere by pressuring the other man. Atanasoff still believed at this point that Iowa State was likely to have filed the patent application. He knew that he himself would not have stolen another man’s ideas, so he didn’t suspect Mauchly—indeed, Mauchly assured him that the principles behind the ENIAC were entirely different from those behind the ABC. A few months later, in August 1944, Atanasoff met J. Presper Eckert for the first and only time, when Mauchly brought him to the gun factory in search of help with quartz transducers. Since Eckert did not have a high security clearance, the two men had to have a military escort, so the visit was brief and unrevealing. Although Atanasoff had agreed to help with the quartz transducers, he didn’t see Mauchly again. It was only later that it occurred to him that quartz transducers could be used in a computer to regenerate memory.
When it was completed, ENIAC was huge. It weighed twenty-seven tons, was eight feet long, eight feet high, and three feet deep. In addition to the 18,000 vacuum tubes, there were 7,200 diodes, 1,500 relays, 70,000 resistors, and 10,000 capacitors for memory storage. It required 150 kilowatts of power, the equivalent of 1,500 100-watt lightbulbs. Because of potential failure of the vacuum tubes, the machine was rarely turned off, but it did malfunction—Eckert said in 1989, “We had a tube fail about every two days and we could locate the problem within 15 minutes.” ENIAC was not a programmable computer—its switches had to be set and it had to be wired to perform its task; if the task changed, it had to be rewired and the switches reset. This could take weeks. The fact that ENIAC was not programmable was a by-product of the speed with which it was built. In his 1943 progress report, Brainerd rejected the added complexity such a feature would introduce—like Atanasoff, he didn’t want to fall into the trap Babbage had fallen into.
As the war progressed in Germany, Konrad Zuse continued to exercise his special genius, which was not just working hard on innovations to his machine, but also making and using all sorts of social connections to circumvent the increasing difficulties of finding materials and developing new ideas. As he began putting together the Z4, he cultivated acquaintances at the telephone exchange who had managed to avoid being drafted into the armed forces by making themselves appear more essential to the operation of German communications systems than they actually were. These “young, energetic, and enthusiastic” friends had access to junk bins, where over and over they turned up parts that Zuse could make use of. And Zuse’s own day job contributed to his understanding of what a computer might do—at one point, he devised a machine for Henschel that calculated optimum wing dimensions for innovative aircraft, a machine that worked fairly reliably for two years. This machine led to another machine designed to “mechanize dial gauge reading.” Although this machine was completed, Zuse had to abandon it almost as soon as he constructed it—he never learned whether it was blown up at the end of the war, or whether the Russian forces captured it. He writes, “Even as I was putting it together, the order came to dismantle the just-completed factory … But I went on working like a madman, driven solely by the ambition to see this interesting machine actually work at least once. Finally, it was created—the first process controlled computer. Even if not a single person had been interested, I had the pleasure of solving a difficult problem once again.”
Zuse and his colleagues began on the Z4 in 1942, building the machine in Berlin in the midst of air raids and fire bombings. On one occasion, Zuse was climbing the stairs in his office building and had just come to a landing when he heard “a crackling sound overhead.” As soon as he ducked into a nearby doorway, the staircase crumbled away. He managed to get down to the cellar and attempted to put out fires with a portable fire extinguisher, but the building burned to the ground anyway. All told, the Z4 had to be moved three times within the city limits of Berlin during the war. Even as Zuse persisted, he writes, “I didn’t always reach the cellar in time” to find safety—sometimes the air raid warnings would sound at just the time he was ready to test some function. But Zuse was dedicated—when he writes about building the Z4 during the war, he suggests that he was more fearful of the computer not functioning than he was of more mortal outcomes:
So, of course, when after weeks or months of work, I know that the time has come for the device to perform without a hitch, then the moment when the start button is to be pressed is especially tense. I always had a pronounced fear of such moments … It takes good nerves to withstand something like this for years on end.
Zuse was not entirely cut off from the outside world, but communication channels were idiosyncratic. At one point, Zuse’s bookkeeper told his own daughter about what Zuse was inventing. The daughter, who worked for German intelligence, responded by reporting that a similar machine was being developed in the United States. Zuse concocted the ruse of sending two assistants to the intelligence offices, where they presented what looked like an official document from the Air Ministry, asking to see the information. They were turned down, but since they had been told which drawer the photo was in, they managed to find it and bring it back to Zuse. The photo was of Howard Aiken’s Mark I. Zuse could not infer many technical details from the photo, but he became further convinced that computer development would have many, many applications in the postwar world. Unfortunately, in Germany, “hardly anyone could imagine commercial applications for our machines. Civilian production would also have been out of the question; it was officially forbidden.”
But Aiken’s Mark I, a machine that looked sleek and elegant (and huge) in the photograph Zuse saw, had a history in some ways as troubled as any of the other machines. Like most of the other scientists working on computers, Aiken joined the war effort (the Naval Reserve) once his PhD was completed. When IBM began building the Mark I (and, subsequently, Mark II–IV), IBM engineers began modifying Aiken’s design. The result was that Aiken became less and less involved with the final design features—the machine was taken over by the institutions that financed it. As the computer approached completion, IBM and Harvard made elaborate plans to unveil it in a joint ceremony. IBM, having spent half a million dollars ($6 million in 2010 dollars) building the machine, was eager to fully share the credit for its design and implementation. Aiken, however, seems to have done something—possibly contacting the press—that shifted the emphasis away from IBM and toward Harvard. Thomas J. Watson, Jr., later said, “If Aiken and my father had had revolvers they would both have been dead.” Hard feelings lingered for years afterward.
Alan Turing is now a famous man—the subject of biographies, papers, an opera, and at least one play, but his work at Bletchley Park breaking the Enigma code did not come to light until the 1970s, and then, at first, only by means of popular books that did not actually mention him, or mentioned him in cryptic ways (F. W. Winterbotham, The Ultra Secret, 1974; A. Cave Brown, Bodyguard of Lies, 1975), or in specialized publications that did mention him directly (Brian Randell, “On Alan Turing and the Origins of Digital Computers,” 1972; Brian Randell, editor, The Origins of Digital Computers: Selected Papers, 1973). Various accounts culminated in an episode about Turing and Enigma in a 1977 BBC series called The Secret War (other episodes concerned radio beams, radar, magnetic mines, and the V-1 and V-2, prototype German cruise missiles). Turing’s genius then captured the popular imagination, but so did his life, which was idiosyncratic, dramatic, and tragically short—he was not only a genius full of charming eccentricities and in some ways a paradigmatic Englishman, he was also an unashamed homosexual. Andrew Hodges, Oxford mathematician and gay activist, published his dense biography of Turing in 1983, which focused equally on Turing’s life and on his work. But there was much more going on at Bletchley Park between 1941 and 1944 than the cracking of the Enigma code.
The essential difference between Enigma messages communicated to German ships and Tunny messages was that Enigma messages were hand encoded, then communicated by radio broadcast, then hand decoded, while Tunny messages, also communicated by radio broadcast, were machine encoded and decoded, therefore not as subject to the human errors that allowed the English decoders to break the Enigma. The Tunny messages were also much more complex. The German army set up a radio network between Ukraine in the east, Brittany in the west, Tunis in the south, and Oslo in the north. Some stations were fixed, but most consisted of two equipment-carrying trucks, one with a sending Lorenz machine, a receiving Lorenz machine, and a teleprinter, the other with radio equipment.
Although in the early 1980s Tommy Flowers was given permission to describe the workings of the code-breaking machine named Colossus that he and his team of engineers built at top speed in 1943, he was forbidden to say what the machine had done or how it had been used in the war. It was only toward the very end of Flowers’s life, when the United States declassified some communications by American liaisons at Bletchley Park that mentioned Colossus and described its function, that the importance of the machines began to emerge (there were ten of them, the first Mark I that Flowers designed and built in 1943, and the nine Mark 2s that were larger and faster, built in 1944). In 2000, the British government finally declassified a long report on Colossus, written by code breakers in 1945, that revealed not only the complexity of Colossus but also its importance—and it was dramatically important.
The job of the Colossus team was the same as that of the Bombe builders—to infer by means of technical and theoretical deduction what the mechanical Lorenz encoding machines were doing and how they worked, and then to build a machine that mirrored that structure. In a teleprinter machine, upon which the Lorenz was based, a long strip of paper about an inch and a half wide passed through a slot the way a piece of paper passes over the roller of a typewriter, short end first. It was advanced by means of a line of tiny sprocket holes about three-fifths of the way between the left edge and the right edge. The pattern of holes standing for each letter of the alphabet and other essential characters according to the Baudot-Murray code, which had been invented by Emile Baudot in 1870, ran across the strip, three holes to the left of the sprocket holes and two holes to the right. The five positions in each row, some punched and some unpunched, represented a letter of the alphabet. For example, the letter M was represented as hole/hole/hole/no/no (or x x x . .) while the letter N was no/hole/hole/ no/no (or .x x . .). A message communicated by a normal teleprinter (or teletype machine, as it was called in the United States) consisted of a long blank strip of paper to indicate that a message was beginning, followed by a strip riddled with lines of holes, the length of which depended on the message, which was followed by another empty strip that indicated the end of the message. Since every letter consisted of five positions (hole or no), a six-letter word, such as “letter.” would consist of six lines. The words of the message ran down the strip: the word “colossus” would have looked like this:
Obviously, such a way of representing letters is time-consuming to generate by hand but easy by machine, easier than Morse code because the machine can punch an entire line at one time.
The job of the Lorenz machine was to take the principle of teletyping and encode the message so that it would be indecipherable except by the target Lorenz machine set to the same key as the originating machine. Since a teletype machine is based on the binary principle that a letter consists of five positions, some of which are punched (“1”) and some of which are not punched (“0”), then the machine used a binary arithmetical process to create the code. In Colossus, Jack Copeland calls this “the Tunny Addition Square” (appendix 3). The letters and symbols in the coded message were passed through the machine and “added” to letters in what was called the “keystream,” or the entirely different order of letters and symbols produced by the machine. The rules of addition were that 0 + 0 = 0, 1 + 1 = 0, 0 + 1 = 1, and 1 + 0 = 1 (note that this addition square is like Boolean algebra, but the values assigned to the results are specific to the rules of the Lorenz machine—it was not a mathematical machine and was not designed to solve math problems). The products of the addition of the coded letters to the keystream letters were systematic, and because the system was binary, if the Tunny receiving machine was set to the same keystream, all it had to do was take the coded message and add the letters and symbols of the keystream to the coded message, and the original message was retrieved. The Tunny Addition Square has 1,024 possible results (just like a base-ten multiplication table has 100 possible results). The more levels or “wheels” the machine employed, the more shifts were possible, and the German encoders employed the twelve wheels of the Lorenz machine in different ways, all of which were organized by headquarters. What the English eavesdroppers soon realized was that part of decoding the message was getting hold of the key (often transmitted between operators by hand) and using it to sift through the messages (transmitted by machine). However, what Turing understood was that with twelve different wheels, the number of possible variations was more enormous than human decoders could manage. Wheels 1–5 operated together (the code breakers called these the “psi” wheels after the second-to-last letter of the Greek alphabet). Wheels 8–12 also operated together (the “chi” wheels, after the third-to-last letter of the Greek alphabet). Wheels 6 and 7 were called the “motor” wheels. Each wheel had a number of positions—wheel 1 had forty-two positions, wheel 2 had forty-seven positions, for example. The job of the code breakers at Bletchley Park was to decipher the patterns in each set of teleprinted letters so that each shift of each wheel could be peeled away to reveal the original message. Intercepted encoded paper tapes were the raw material that Colossus had to process. Uncovering the shift pattern of one of the encoding wheels of the Lorenz machine was the key—once the position of the first wheel was ascertained, the positions of the next wheels became progressively easier to ascertain through Boolean logic. But while Enigma had three wheels, and then four, which was difficult enough, the Lorenz machine’s twelve wheels hugely enlarged the number of possibilities that had to be tested. And though sometimes with Enigma, the German operators encoding and sending the messages made mistakes that gave away the pattern, the mechanization of the Lorenz encoding process gave rise to fewer human errors, which was a large part of the reason Tunny was more difficult to decode.
In order to gain some idea of the work Colossus had to do, let’s imagine a message of five hundred holes and spaces representing one hundred letters (a very short message). It was the job of German intelligence officers to designate the positions and of the Lorenz operators to set the positions. Until the summer of 1944, the position of the psi wheels was set monthly and the chi wheels quarterly, then monthly. The motor wheels were set daily. As the war heated up in 1944, the positions of all the wheels changed daily.
The Dollis Hill communications research laboratories were located about eight miles northwest of central London, in an area that had originally been farms, then the estate of a politician who was a friend of William Gladstone and who had served as governor-general of Canada and lord lieutenant of Ireland. As close as it was to central London, the area retained its rustic feel into the twentieth century. But by the First World War, the team designing the Liberty tank, Mark VIII, was based there, and in 1921 the English government established the Post Office Research Station there. By 1933, a large brick factory and offices had been built, and at the beginning of World War II an underground bunker called Paddock was installed (though Churchill didn’t like it and wouldn’t stay there). The parts of the Colossus were shipped to Bletchley Park (about an hour’s drive farther northwest) and assembled there.
It was Tommy Flowers who conceived and built Colossus at Dollis Hill, where he had worked since 1926. Even though because of his prewar vacuum-tube experiment Flowers knew how much faster the tubes were at such work, in 1943 he could not at first persuade the authorities at Bletchley Park to try the new technology. He decided to construct a prototype on his own, commandeering a post office factory in Birmingham to make the parts. He had a sixteen-hundred-tube processor by the end of 1943 but saw immediately that though it worked, it was not fast enough, and he began on an improved version in February 1944. He was told that the machine had to be installed at Bletchley and functioning by the first of June, the planned date for the invasion of Normandy by the Allied forces. He succeeded. According to Jack Copeland, “Despite the fact that no such machine had previously been attempted, the computer was in working order almost straight away and ready to begin its fast-paced attack on the German messages.” Not long before he died, Flowers did write enough about the history, the purpose, and the features of Colossus so that we may understand its main features:
Colossus was a special-purpose machine designed primarily to perform processes devised by Bletchley Park for discovering the settings of the code wheels made by the [German] machine operators before the messages were sent. Much of the Colossus was an electronic analogue of the Lorenz Tunny machine. Bletchley Park also eventually found ways of using the machine to discover the Tunny wheel patterns when they were routinely changed. (Colossus did not itself decode intercepted messages. This was done by other machines, specially modified teleprinters, also known as Tunny machines.)
The Colossus operated on two data streams simultaneously—one was the strip of paper from the teleprinter, carrying the message, and the other was a data stream that mimicked various wheel combinations that a Lorenz machine would use. The strip of paper carrying the hole pattern that was the message was made into a loop, then the loop was passed over and over through a photoelectric reader that registered hole or no—each recognition registered as an electric impulse to the logic unit (the “processor”—the part of the machine that eventually would be made up of 2,400 vacuum tubes). Each pass of the loop through the scanner included a blank section that defined the beginning and the end of the message. The tape passed through the scanner over and over “until every possible combination of digits” that appeared at the beginning of the message had been read—once the beginning of the message had been worked out, the rest of the message could be decoded. The electric impulses that passed through the holes in the tape registered on a counter; the code breakers soon discovered that a scan that did not reveal a message always contained fewer impulses than a scan that revealed a message—that is, the word “colossus” contains eight letters, and so, eight lines of holes and nos; in “colossus,” there are eighteen holes versus twenty-two nos. No eight-letter word could contain, say, three holes and thirty-seven nos. According to probability, every eight-letter word had to contain more than a certain number of holes, so Colossus was set to throw out results that contained fewer than that number. Colossus allowed the code breakers to concentrate on only the strips of letters that were more likely to resolve into the actual message.
One flaw in the Lorenz machine, as a system of rings, was that somewhere in every message was a spot where the wheels returned to the start position. This meant that the encoding, though large and complex, was not perfectly random. Since the machine that the Germans were using was made of wheels and gears, it, according to Flowers, “generated and processed numbers” rather slowly—five every second. Colossus, because of the vacuum tubes, was a thousand times faster, its speed limited by the passing of the paper strip through the reader, not by the speed of the vacuum tubes. Since the Colossus was essentially a sorter, Flowers wanted it to sort as quickly as possible—and five thousand times per second was not fast enough, so a shift register was invented that read, counted, and kept track of five different readings of the holes each time the tape was passed through the machine. Colossus read and counted the holes so quickly that the code breakers could usually narrow in on the telling spot fairly quickly. Once they had done that, the pattern of the code was revealed, and the message could be broken. Colossus also had a mechanism for detecting and discounting spots where a message might have been incorrectly received.
D-Day was set for June 1, 1944, but as it happened, the invasion was postponed because of the difficulty of moving troops and materials in bad weather. According to Andrew Roberts, when the chief meteorological officer, James Stagg, was handed the list of weather requirements that suited each faction of the invasion force, he said, “When I came to put them together I found that they might have to sit around for 120 to 150 years before they got the operation launched.” But there were concerns other than weather—principally the question of what the Germans thought the Allies were planning. On June 5, Eisenhower was interrupted in a staff meeting by a courier bringing the first Colossus-decoded German communication from Bletchley Park. Flowers writes, “Hitler had sent Field Marshall Rommel battle orders by radio transmission, which Bletchley Park had decoded with the aid of the new Colossus. Hitler had told Rommel that the invasion of Normandy was imminent, but that this would not be the real invasion. It was a feint to draw the troops away from the channel ports, against which the real invasion would be launched later. Rommel was not to move any troops. Eisenhower read the paper silently, then announced, ‘We go tomorrow.’ And on the morrow, 6 June, they went.”
With the help of Colossus, the decoders at Bletchley Park then decoded Hitler’s subsequent messages to his armed forces and preempted his attempt to foil the invasion. According to Flowers, “The result was a defeat of the German Army so overwhelming that the Allies were able to sweep rapidly eastwards across France.” According to Roberts, Eisenhower also remarked to his staff, “I hope to God I know what I’m doing.” But Allied intelligence and counterintelligence worked so well that “even up to 26 June half a million troops of the German Fifteenth Army stayed stationed around the Pas de Calais, guarding against an invasion that would not come.”
Flowers felt that he was the pivotal man in the success of Colossus because of his familiarity with vacuum tubes. He writes, “If I had … spent the war interned in Germany, Colossus would not have been built, because there would have been no one at Dollis Hill with sufficient knowledge of the new technology to make it. If Dollis Hill had not made Colossus, some other organization may have made something similar, but we now know that none could have done so by D-Day. Those chance events changed the course of the Second World War. If they had not, history would now record the devastation of Europe and a death toll much greater than actually occurred.” One key feature of Colossus’s success was that Flowers, like Eckert, realized that the vacuum tubes, which were seen as unreliable when he first began to use them, were much more likely to fail at the moment of thermal shock when being turned on. For the fifteen months that Colossus was at work, a machine was only turned off if it was malfunctioning.
Flowers and his fellow inventors were not only proud of their machine, they were thrilled by it. The engineers who authored the report on Colossus at the end of the war (the report that was declassified in 2000) wrote:
It is regretted that it is not possible to give an adequate idea of the fascination of a Colossus at work; its sheer bulk and apparent complexity; the fantastic speed of thin paper tape round the glittering pulleys; the childish pleasure of not-not [sic], span, print main header and other gadgets; the wizardry of purely mechanical decoding letter by letter (one novice thought she was being hoaxed); the uncanny action of the typewriter in printing the correct scores without and beyond human aid; the stepping of the display; periods of eager expectation culminating in the sudden appearance of the longed-for score; and the strange rhythms characterizing every type of run: the stately break-in, the erratic short run, the regularity of wheel-breaking, the stolid rectangle interrupted by the wild leaps of the carriage-return, the frantic chatter of a motor run, even the ludicrous frenzy of hosts of bogus scores.
Flowers invented Colossus, but he also gave credit to Alan Turing for his contribution. At a conference in 1980, Flowers saw a young man reading the book that grew out of the BBC series The Secret War. The two struck up a conversation, and Flowers recalled, “You’d be working on a problem and not able to solve it, and sometimes someone would look over your shoulder and say, ‘Have you tried doing it like this?’ and you’d think, ‘Of course, that’s how you do it!’ With Turing, he’d say ‘Have you tried doing it this way?’ and you’d know that in a hundred years you would never have thought of doing it that way. And that was the difference.”
In the course of the eleven months between D-Day and the German surrender in May 1945, the General Post Office built and the intelligence services made use of ten Colossus machines. According to Flowers’s obituary by Alan Blannin in the Daily Telegraph, “At the end of the war, all but two of the Colossus machines were destroyed. Flowers was ordered to destroy all evidence that they had ever existed. The two surviving machines were taken first to Eastcote, west London, the first home of the new Government Communications Headquarters, and then to its present base at Cheltenham, where a Colossus was still operational in the early 1960s.” Flowers, however, did not have access to them.
The code breakers at Bletchley, even with ten Colossus machines, did not break every message, but the Germans did not expect them to be able to break any messages, and so they continued to use the Lorenz machine for high-level army communication even after they should have deduced from the failure of certain operations that something was wrong—in fact, Thomas Flowers worried about being too successful and thereby undoing all of his own work. There were other machines and other methods of encoding that the Germans used and the English did not break, but since the Germans chose to use the Lorenz machine for army communications at a time when the war was an army war across France and into Germany, Colossus was, in the eyes of its creators and others, the key to victory. It was this euphoria that led Thomas Flowers to accept the destruction of the Colossus machines and the ban on discussing either how the machines worked or what they had done between June 1944 and May 1945. The obituary in the Telegraph pointed out a further irony: “Flowers received very little remuneration from the government for his invention … barely sufficient to pay off the debts that he had run up while developing Colossus.” According to most sources his insufficient remuneration amounted to about £1,000 (some $40,000 in 2010 dollars, or about five times what Atanasoff had been granted for the development of the ABC).
Charles Babbage, 1791–1871, inventor of the Difference Engine and Analytical Engine, analog computing devices. (Photograph courtesy of the Charles Babbage Institute, University of Minnesota, Minneapolis)
A section of Babbage’s Difference Engine, showing rods and gears. (Science Museum/SSPL)
Vannevar Bush with his Differential Analyzer, 1931. (Courtesy MIT Museum)
John Vincent Atanasoff, around the time he completed his PhD at the University of Wisconsin. (Iowa State University Library/Special Collections Department)
Atanasoff in the 1930s, teaching at Iowa State College. (Iowa State University Library/Special Collections Department)
The physics building at Iowa State College. Atanasoff and Clifford Berry built the ABC in a corner of the basement. (Iowa State University Library/Special Collections Department)
Clifford Berry, 1918–1963, standing with the ABC in 1942. (Iowa State University Library/Special Collections Department)
An undated schematic of the ABC, prepared for a campus exhibition at Iowa State University. (Iowa State University Library/Special Collections Department)
The ABC in May 1942. (Iowa State University Library/Special Collections Department)
One of the ABC’s two electrostatic memory drums, the only surviving part of the original machine. (Courtesy of U.S. Department of Energy’s Ames Laboratory)
Konrad Zuse’s Z1 computer, built in his parents’ Berlin apartment c. 1936.
(Courtesy of Horst Zuse)
Konrad Zuse, 1910–1995. (Courtesy of Horst Zuse)
Alan Turing, 1912–1954, upon his election as a Fellow of the Royal Society in 1951. (© National Portrait Gallery, London)
Bletchley Park staff at work on deciphering codes, Hut 6.
(Bletchley Park Trust Archive)
A Lorenz SZ42 Schlüsselzusatz cipher machine on display at Bletchley Park. (Bletchley Park Trust Archive)
Thomas Flowers, 1905–1998. (Bletchley Park Trust Archive)
Colossus at work in 1943; note paper tape.
(Science Museum/SSPL)
Aiken’s Mark I analog device in use, 1944.
(Courtesy of the Computer History Museum)
John Mauchly, 1907–1980 (left), and J. Presper Eckert, Jr., 1919–1995 (right), with Major General G. L. Barnes, 1944. (University of Pennsylvania Archives)
ENIAC in 1946—Eckert stands front left, while Mauchly is by the column. (University of Pennsylvania Archives)
John von Neumann with EDVAC in 1952; note Williams tubes along the bottom of the machine. (Alan Richards. photographer. From the Shelby White and Leon Levy Archives Center, Institute for Advanced Study, Princeton, NJ, USA)