24

Computing machines

Two United States army officers arrived at the Sydney waterfront in March 1943, asking the wharfies questions about shipments from California, and searching around through stacks of crates. They found what they were looking for — several boxes that had been sitting on the dock for several weeks — loaded them into a truck, and drove away. The two Americans were Major Larry Clark and Lieutenant Richard C. Frazier, and the boxes they had located contained the parts for electric calculating machines built by an American company called IBM. The machines were needed in Brisbane at a secret organisation called Central Bureau.

They transported the boxes to Brisbane and offloaded them into the garage behind the main house at 21 Henry Street, Ascot. There, the boxes were opened and their contents examined. There was some weather damage and some pieces were corroded, but nothing that looked too serious.1 The bigger problem was that the IBMs needed to be assembled, and there were no assembly instructions.

Inside the boxes were parts, neatly stored in envelopes with numbers written on them, but there was no parts list to tell them what the numbers meant. Luckily, Frazier had worked with IBMs in Washington. He recognised many of the parts, and could explain to the others what they did and where they fitted into the machines.

Another clue was a photo of an individual machine, as it would look completely assembled, on the outside of one of the boxes. With the photo on the box, Frazer’s recollections, and using their combined expertise, they were able to assemble the IBMs. Then they discovered that the Australian power grid runs at 240 volts, while the IBMs were designed for the American grid, at 120 volts. They needed step-up transformers, which they ordered to be custom-built in Melbourne and shipped to Brisbane, from where they were unloaded into the garage behind the building.2

An older IBM, a tabulating machine, was already operating in a commercial business in Brisbane. Upon Central Bureau learning of its existence, it had been commandeered, loaded onto a lorry, and installed at Nyrambla. To house it, the roughly built garage walls were insulated with two layers of caneite (compressed boards of sugar-cane fibre), and a large air-conditioning unit was installed.

There were now 12 new machines installed. When they were configured correctly for a code that had been broken and for which the codebook was known, the punch-card operators could set up cards with the numbers from an incoming message, and the machine would read the card, strip the additive, decode the groups, and print out a Katakana version of the text.3

A punch card, a thin piece of cardboard with holes punched in it, was the primary way of providing input. There were no electronic keyboards or touchpads; the computer could only respond to the patterns of holes in a punch card. As it was fed in, brushes would pass over the card, triggering switches if they dropped into a hole and thereby relaying the contents of the card. Levers triggered by the switches would flick the cards into piles according to the configuration of holes. The machines made a lot of noise.

The air conditioning was necessary, not for the machines themselves, but for the cards. If the cards got damp, the machines would not read them correctly. The cards had to stay cool and dry.

WAAAFS — personnel from the Women’s Auxiliary Australian Air Force — were put to work on it. Eve Scott, who had enlisted in the WAAAF, started out as a store clerk before being abruptly moved to Central Bureau and a secret world of ciphers and machines.

‘We were sent to 21 Henry Street, Ascot, Headquarters at Brisbane,’ she said. ‘We were only there a short time, and all of a sudden we were moved from there to Oriel Park, Ascot in the fire station. And when I got there I was so scared of these machines. I didn’t want to learn.’4

An American officer patiently helped her overcome her trepidation, and soon it was ‘head down, tail up’, and she was part of a workforce that processed up to one thousand Japanese radio messages a day.

Before the computers arrived, everything was done with paper and pencil.

Interceptors at the field stations transcribed radio messages by hand; the messages were transcribed into Kana by hand; and by hand they were sorted, decrypted, and translated. Abe Sinkov had seen a better way of doing this in Washington. There, his colleague and mentor William Friedman had pioneered the use of machines that could churn through computations many times faster than humans.

At the outset of the war Sinkov and Friedman, the founder of cryptanalysis in America, and the man responsible for breaking the Purple machine, had approached the founder president of IBM, Thomas J. Watson Sr., for help.5 Watson’s company, IBM (which stood for ‘International Business Machines’) manufactured business equipment such as typewriters, and collating machines of various kinds, and had a reputation for innovation and the commercial application of technological advances. It had grown from around 6000 employees and $19 million in revenue ten years earlier to a company with more than 18,000 employees and $86 million in revenue in 1942. This was due to innovations such as the first commercially successful electrical typewriter, the invention of an automated carriage return on typewriters, and the first automatic marking system using cards, known today as ‘fill in the bubble’ answer sheets.

Watson put all the resources available in the company at the army’s disposal, and chose Stephen W. Dunwall, a product developer at the IBM factory at Endicott, New York, to work with Friedman on developing and running IBM machines in the code-breaking effort.6 They brought in IBM punch-card machines, and adapted them to new uses. The machines could strip additives off code rapidly and in bulk, and look for repeating patterns in the resulting layer of code beneath the additive. Repetition might mean that the same message had been encoded twice with different additives. In this way, they could try thousands of combinations until they hit one that looked promising.

Now in Australia, Sinkov knew what he needed: he was up to speed on the use of such machines by the American cryptographers; Central Bureau could benefit from the power and efficiency of these machines; and IBM was producing the best model for the job, the NC4. It was a no-brainer. Sinkov ordered a shipment of IBMs to be transported to Brisbane from America.

In the meantime, Sinkov’s colleagues at Special Intelligence Service (SIS) were willing to run calculation jobs for Central Bureau on their IBMs in Washington. SIS was expanding rapidly, and had recently relocated to a building called Arlington Hall in a girl’s school that the US army had acquired for the war effort.

Sinkov sent them a batch of 2000 messages in a code known as the 666 code, which was a three-figure Kana code, with a larger sample of messages than most of the other tactical codes. It was being used by an unknown Japanese unit in New Guinea. Arlington Hall’s machine technicians tried what they called a ‘brute force’ attack. This basically involved getting the IBMs to churn through every possible encoding, looking for a meaningful message underneath. It did not work, because the messages did not have enough overlapping indicators.

This did not dampen Sinkov’s enthusiasm to acquire some IBMs for his own use at Central Bureau. After several months, for some reason, they still had not arrived. He sent an enquiry to Washington asking when they would be shipped, and was informed that the IBMs had been sent months before.

Doug Pyle, previously a Kana operator, was for a time an IBM mechanic after a dreary stint working as a clerk for Eric Nave. Having returned from the field due to ill health, he was stationed at a desk in the hallway at Nyrambla outside Nave’s office, poring over pages of code under a dim light. He complained to Eddie Kelson that it was ruining his eyesight, and added that he would like to work with the IBMs, as he was ‘good with machines’. Pyle was immediately appointed as IBM mechanic, under the tutelage of Van, an affable American with an endless store of ribald stories.7

The IBMs needed to be kept cool, and so large industrial-cooling systems were installed, but they were so noisy that the neighbours complained. The machines were relocated to nearby Ascot Fire Station.

The machines were used for simple tasks at first, like sorting words. A basic trick in code-breaking, if you can do it, is to find the high-frequency code words. This is a useful first step in finding out what those words mean. Sorting words meant creating punch cards for every word in every message — a laborious task. By the time the IBMs arrived at Nyrambla, Sinkov’s colleagues at Arlington Hall, under the direction of Solomon Kullbach, had devised more efficient methods that involved much less labour.

The tabulators could do other useful tasks. They could sort by call sign, for instance, so that all messages coming from a particular enemy unit could be grouped together. Or, alternatively, they could sort a large pile of messages by indicator values, such as all messages using the same set of additives to encipher the code words.

The IBMs, marketed and sold as sorting machines, were not computers in the modern sense. But the code-breakers, over the course of the war, found ways of making them do more and more complex tasks. Army engineers built ‘add-ons’, plug boards that could be installed onto the machine as a kind of program.8