Hidden Communications

World War II was rife with improvements to computer technology, and much of it focused around two key areas - ballistics calculations, somewhat unsurprisingly, and cryptography. Cryptography is, simply put, the art of hiding information in plain sight - somebody wishes to send information to somebody else, without anyone but the intended recipient also being able to read it. You remember those little secret codes you came up with as kids? That's cryptography, though cryptography in the real world tends to be way more complex.

The German war effort used two cyphers, Enigma and Lorenz. Enigma was used by the military for day-to-day use, and Lorenz was used by the officers and government for important messages. Needless to say, capturing these communications was of huge importance to ending the war. Thus, a significant effort was put into breaking both cyphers.

Enigma was arguably simple. It was available commercially in the early 1920s, then developed into a more complex military version. It worked by passing letters through three sets of scrambled rotors. The first would click one position over every message, the second would move every full rotation of the first, and the third every full rotation of the second. The letter would then pass back through the rotors once more, before illuminating a bulb. These rotors could be changed for extra variance, and the military added a "plug board", allowing letters to always be swapped for one another before and after the rotors by connecting wires.

This was more complicated than any cypher seen to date. So much so, that the standard military issue Enigma used in World War II had approximately 159 quintillion different positions. Thus, breaking this by hand was rather unlikely.

Polish Ingenuity

Poland had a keen interest in breaking Enigma, in case of another war. This effort was begun very early in the 1930s, by a team of 3 mathematicians. Marian Rejewski, a Polish mathematician, investigated the commercially available Enigma machine, and from this was able to produce a mathematical model. This was then used by Henryk Zygalski to develop a method to decrypt Enigma, known as Zygalski sheets.

Later, in what historians believe to be late 1938, Rejewski invented the "bomba kryptologiczna", or cryptologic bomb. This was a machine that could, in effect, churn through possible combinations of Enigma settings, and could extract the original settings in approximately two hours, letting operators then decode the original message. This came as the Germans began ramping up the number of encrypted messages sent. Decrypting these gave a very clear indication - they were preparing for another war.

In December 15th, 1938, the German war effort introduced two new rotors, 4 and 5, dramatically increasing the complexity of the cypher. The result of this, another very clear indication that war was afoot, was that in July 1939, Poland shared the bomba with Britain and France before it was too late. Shortly thereafter, on September 1st, 1939, Germany invaded Poland, leading Great Britain and France to declare war on Germany.

British Developments

Located at Bletchley Park, Britain had deployed a team of mathematicians and cryptanalysts selected from among the elite, many of whom came from Oxford and Cambridge universities. The location was selected for ease of access, being roughly central to both Oxford and Cambridge, and with a rail line to London.

Among this team was the father of computing, Alan Mathison Turing, who designed the original Bombe machine in 1939, a design that was built upon by Gordon Welchman in 1940. This worked by mounting a "known plaintext" analysis - one took part of a message predicted to be in the message to decode, and attempted to find its location using known properties of the Enigma. From this, the Bombe could be used to determine the machine settings.

This allowed decryption of army and air force messages, however the German navy used 3 more wheels, hereunto unknown. Turing managed to allow parital decryption of this traffic in 1941, following the capture of a U-Boot.

Lorenz

Lorenz was an altogether more complicated machine, relying on modulo-2 addition to encypher text, representing letters in binary using a system called the Baudot code. This proved nearly impossible to decode for the British, until August 1941, when a Lorenz operator sent a message twice, on the same machine settings, with small differences in each message. This allowed John Tiltman, another Bletchley Park mathematician, to crack the code, recovering both messages.

Tiltman passed his work on to Bill Tutte, a chemistry graduate from Cambridge University. Tutte used this to work out the structure of the machine, which was later used to create a machine, Tunny, which emulated the Lorenz. Tunny was designed and produced by Frank Morrell, and took about a month to decode each message. This was a large step up from never, but still presented a large issue, as by then the strategic importance of the message may have been lost simply due to it no longer being relevant.

British Rube Goldbergs

Max Newman, another Bletchley analyst, worked with a team of electronic engineers to design a machine later dubbed the Heath Robinson, so named after a cartoonist who designed vastly complex machines to achieve seemingly simple tasks, much like Rube Goldberg. The Heath Robinson resembled one of these contraptions, and was hence dubbed.

The Heath Robinson automated part of the process of determining the settings of the Lorenz machine, by using various statistical techniques developed by Tutte. It used paper tape to encode both message and wheel settings, and, in a world first, did not use electromechanic components for computation. Rather, this was one of the first uses of thermionic valves (vacuum tubes) for computation. Up to this point, they were almost exclusively for radio equipment.

Valves were significantly more reliable than relays and uniselectors, having no moving parts, however suffered the problem that turning them on and off repeatedly eventually caused them to burn out, much as an incandescent lightbulb does (indeed, an incandescent lightbulb is, in fact, a thermionic valve itself, just a very simple one performing no electrical function outside of "emit light". All valves emit some level of light when turned on, and incandescent bulbs take advantage of this fact).

A Valve Computer

Shortly thereafter, an engineer at the Post Office, Tommy Flowers, was drafted in. Flowers criticized the unreliability of the Heath Robinson, and several aspects of its design. Taking inspiration from it's valve elements, however, he went on to build a much more complex machine, capable of computing more reliably and removing electromechanical parts entirely, other than a single tape reader (actually, two, but they did the same job and were only there so one could be run while operators loaded the other).

This machine was known as Colossus. It eliminated an entire paper tape reel the Heath Robinson relied on, and did away with all relays and uniselectors, meaning significantly more reliable operation and much faster computation. This drew the criticism, as mentioned above, on valve failure upon power on, leading him to remark, "the trick is not to turn them off". Thus, the Colossus was never powered down.

By December 1943, the Mark 1 Colossus was shipped to Bletchley Park, and set up in F Block. It was operational and in January 1944, and, on February 5th, 1944, broke a German message that had not yet been cracked. Flowers was present and observed this, remarking in his journal on the day, "Colossus did its first job. Car broke down on way home." The operation of Colossus reduced the time needed to break Lorenz cyphers from weeks to mere hours, and provided crucial tactical information to British and American command centers. It was estimated that this reduced the war by 2 whole years.

The Colossus was built to be as generic as possible, meaning that as new techniques for breaking the Lorenz were found, the machine could make use of them with a minimal effort. This then lead Donald Michie and Jack Good, in February 1944, to discover a new technique to break wheel patterns quickly. Flowers was instructed, as a result of this work, to develop a new machine, the Colossus Mark 2, which had a control panel to control this particular process.

The Mark 2 was delivered to Bletchley Park on May 4th of that year, though was plagued with unreliability. It was promised to work on June 1st, however Flowers and his team were working late into the night on May 31st, diagnosing unreliability after unreliability. As midnight passed, the team retired to bed, leaving W. W. Chandler, one such engineer on the team, to work alone on the machine. The following morning, the Mark 2 was operating correctly, however a burst radiator pipe resulted in a puddle of water making its way towards the Mark 2. The navy's solution to this was for the wrens operating the machine to wear rubber soled boots while working on the machine, insulating themselves from the high voltages potentially running through the puddle.

By the end of 1944, 7 Mark 2 Colossi were in operation at Bletchley Park, dramatically aiding the D-day assault and ensuing front as the British faught to end the war. By the end of the war, 10 were in operation, with an expected delivery of an 11th shortly thereafter, which was, of course, cancelled.

Postwar Effect

As a result of the war, Flowers was bankrupt, Newman turned down an OBE, and Turing recieved one. Prime Minister Sir Winston Churchill ordered the Colossi dismantled, and the codebreaking program was kept strictly secret despite the end of the war.

Additionally, the Government Code and Cypher School, the organisation overseeing the operation of Bletchley Park and several other codebreaking sites, became the Government Communications Headquarters, or GCHQ, which still operates today. Upon setting up and moving to Cheltenham, GCHQ took several Collosi before they were destroyed, making use of them until 1960, with one, dubbed Colossus Blue, being retired in 1959. It is not known what these were used for originally, however in the final years of their operation, they were used for training purposes.

It is known that Jack Good worked on one of these machines, on a program for the NSA, the US equivalent of the GCHQ. The program he worked on remains classified, but the NSA were prepared to commission a new machine. Implementing this on the Colossus earned Good significant prestige within the NSA, leading him to remark that "Golde told me that one of his friends who visits NSA told Golde that I am 'regarded as God' there." This demonstrates the generality built into Colossus.

Technologies of the Time

Each of these technologies shared a commonality - they were wired for one specific job. Colossus was an outlier here, though only minorly so, in that it could be rewired to perform a different task. No, not reprogrammed, rewired. Storing "software" as data was a quirk of history that had been forgotten about, though would soon be rediscovered.

Also visible at BeaR&D

Wartime Effort, from the Other Side

Most people, when they think about WWII, don't think about the German perspective, and when they do, it's usually considered from the point of view of "evil enemy nation who did some stuff." Today, I shall consider quite a different side to the German war effort - the technological advances that came from it, specifically within the realm of computation. Of course, this is a vast topic, so only a small part of it shall be considered in this post.

Mechanical computing was used a lot during WWII, and in the years leading up to it, predominantly by the military, on both sides, though academic applications were also quite prevelant. One man was particularly famous for his mechanical, and elecctromechanical, machines. This was Konrad Zuse, a serial computer engineer who invented many such machines.

The first, the Z1, at the time known as the V1, was developed and built between 1935 and 1938. Renamed to differentiate it from a weapon used during the same time period, and destroyed during a bombing run on Berlin in 1943, the Z1 was the first of its kind, in many respects - it significantly expanded the capabilities of computers at the time. The computer used binary to represent numbers, and calculated using Boolean logic (operations on the 1s and 0s), simplifying computation and storage significantly.

Additionally, the computer could represent and calculate with real numbers (decimal numbers, as opposed to just whole numbers), allowing the machine to be far more effective at scientific computation. No other computation device in the world so far had these features.

Alas, the Z1 was not reliable in operation. It relied almost exclusively on metal strips, rods, and other such mechanical parts to represent the individual values and computational elements, which would regularly stick, as the parts were often not synchronised with one another. In a classic display of Silicon Valley culture, many years before Silicon Valley was even dreamt of, Zuse constructed the Z1 in his parents' apartment, using funding from friends and family to do so.

Later, Zuse built upon the Z1, replacing the arithmetic and control machinery (which we'd now refer to as a CPU or a processor) with relays taken from telephone exchange equipment. This vastly improved the reliability of the machine, now dubbed the Z2. This work was based on the work, and advice, of Helmut Schreyer, who while studying for a Doctorate at the Berlin Institute of Technology in 1937, worked on memory built from Boolean logic.

Schreyer then worked with Zuse to build further upon the design, arriving at the Z3 machine. Composed entirely of relays, the Z3 was designed for use by the German Laboratory for Aviation, as part of the war effort, where it was used in research on flutter, a problem in aerodynamics and therefore in aircraft design. For this reason, the machine was kept very secret by the Reich.

The Z3 is notable for more than just its political position, however. It added exception values to its computation, as well as +/- infinity, and undefined. This meant that any errors during a calculation could be passed through the calculation, and rather than giving a random value at the end, clearly presented as an error. The easiest example to point to would be if the machine attempted to divide by 0.

Additionally, the Z3 is considered the first machine to be Turing complete. The technical definition of Turing completeness isn't too important here, however this was a major milestone - what it effectively means is that the Z3 was the first general purpose computer. It could be used to perform any calculation, as opposed to fixed equations. The way it did this was unorthadox, and is no longer used, requiring every possible calculation to be done and the unneeded ones discarded, however it was possible nontheless.

Zuse KG, the Z4, and Plankalkül

The Z4 was an expansion of the Z3, once again, simplifying it significantly in places and adding new capabilities, such as abstract references (data could now be named, as opposed to needing to recall its location manually), several instructions such as trigonometric functions and square roots, and the ability to start running a different piece of code, and return later, a technique known as subroutines. It also added the ability to change which instructions were run based on some condition (if this do that, otherwise do something else).

More notably, however, the Z4 was the words first commercial digital computer. Originally, as with the Z3, it was intended for use by the German war effort, however the construction of the original machine was completed days before the war ended. In February 1945, to prevent the machine falling into Soviet hands, the Z4 was moved from Berlin, and relocated to Goettingen, where it was completed (and later, moved again to Hinterstein for the same reason).

Zuse founded Zuse KG in 1949, a company through which the Z4 was commercialised. The first customer was ETH Zurich, who operated the machine from 1950-1954. For a while, the Z4 was the only digital computer in continental Europe.

After selling the Z4, Zuse KG then went on to sell several more machines, of several different designs - the Z5, the Z11, and the Z22. Each added new capabilities to the machine, such as new instructions,new devices, and new technologies such as vacuum tubes.

Zuse, and Zuse KG, aren't notable solely for the Z-series of computers, however. In 1942, Zuse developed Plankalkül, one of the first programming languages, for use on his computers. This work was not published for a long time, due to the effects of the war, however Zuse continued to work nontheless, working on a chess program, and a book about the language, which was never published. The language is remarkably different to nearly every other programming language, and has been compared to APL. It draws a significant amount of inspiration from mathematical notation, and builds upon it, generalising it in the process.

Also visible at BeaR&D

 

Reinventing the Punched Card

A pattern has emerged, as we've seen, when it comes to the invention of computer technology. A job is manual, labour intensive, and repetitive, somebody notices this, and promptly automates it. This is precisely waht Herman Hollerith set out to do in late latter half of the 1880s, after seeing the US 1880 census take 8 years to process.

Hollerith then effectively rediscovered the idea of a punched card to store data, after seeing similar techniques used on train tickets. Along with this, he built a machine, which he called a tabulator, which could count the amount of cards that met a certain criteria - for example, one could set the machine up to count the number of employed males over the age of 50, by having one hole assigned to gender, one hole assigned to employment status, and a group of holes assigned to age band. This was used, to great success, in the 1890 US census, dramatically shortening the time it took to process the results.

The next development Hollerith devised was an integrating tabulator - a machine that could be set up not only to count cards that met a certain criteria, but to add numbers represented in certain combinations of holes. He sold this under the Tabulating Machine Company, which soon thereafter became the Computing-Tabulating-Recording company, CTR for short, along with a series of other companies. This machine was used in the 1900 US census.

In the years that followed, several smaller developments were developed, such as printing the results onto paper rather than on a numerical readout, printing summary cards, more complicated calculations and criteria, and so on. In parallel, several other machines were developed to work with tabulators, such as card sorters, collators, and punches.

Shortly thereafter, CTR changed it's name to International Business Machines, incorporated, initialised IBM. IBM tabulators would be used for decades following this, ending production in 1976. Computers slowly replaced tabulators from the late 1950s onwards, and following the end of the IBM 407, the tabulator era was over. During this time, the capabilities of tabulators were steadily advanced, allowing more and more advanced selection, sorting, and arithmetic calculations on the cards that were fed in, as well as faster operation and improved use of other devices such as printers.

Scientific Revolution

Tabulators were among the first automated computation devices used in science. While mechanical calculators were in use for decades, or indeed centuries by some counts, previous to this, the use was driven manually - a calculation was input, and the calculator would calculate it and display the result.

The tabulating machine, however, was something altogether more revolutionary. Now science had the ability to be automated - one could encode a set of data on a set of cards, program the machine with the correct calculations, and leave the machine to run, potentially for hours, performing calculation after calculation while a researcher performed other duties.

This was achievable primarily due to the basic purpose of a tabulator. Invented to perform statistical calculations for census data, nothing was in place to limit the statistics solely to census data - a punched card could just as well represent data on a person as it could a molecule or astronomical entity, and the calculations were practically identical.

One of the first people to use tabulators in this manner was Leslie Comrie, an astronomer in New Zealand. He used a tabulator to compute lunar position using Fourier synthesis, achieving dramatically higher accuracy than was possible by hand calculation. This work, in 1928, was unheard of in the scientific community, and eventually lead to his suspension from the Nautical Almanac Office in 1936.

Comrie later on founded Scientific Computing Service, Limited, the worlds first private company offering scientific computing services. The company offered services to the war effort, calculating balistics tables for the Department of Defence.

Comrie's work was later built upon by Wallace John Eckert, another pioneer in scientific computing. Eckert, in 1933, developed work based around the principle of chaining tabulators together, to achieve more complex equations. He used this method, with an IBM 601 donated by Thomas J. Watson, the president of IBM, to build upon the results of Comrie's lunar position, improving the accuracy. In doing so, the laboratory Eckert worked in was renamed the Thomas J. Watson Astronomical Calculating Bureau.

Eckert then went on to work with the US Navy, automating the production of navigation tables through the use of tabulating equipment in order to reduce human demand on the process. This work would be continued by the team on the Manhatten Project, spearheaded by Nicholas Metropolis and Richard Feynman.
Later, at Columbia University, Eckert commissioned from IBM a large computing machine, known as the Selective Sequence Electronic Calculator, or SSEC, bringing scientific computing into the era of computers, rather than tabulators. This work was completed in 1947.

Meanwhile, Feynman and Metropolis worked on the automation of computing on the Manhatten Project, once again using IBM equipment. Details on the use of this computation equipment are few, due to the nature of the project itself. Both Feynman and Metropolis were influential in computing in later years as well, Feynman with work in supercomputing and quantum computing, and Metropolis working on teams to develop bespoke computers as was the default in the 1940s, as well as pioneering the Monte Carlo method of simulation, a technique still used today for scientific computation.

Also visible at BeaR&D

 

Looming Changes

In the early days of fabric weaving, it took two to operate a loom - the weaver, who would manipulate the thread, and the draw boy, who would operate various mechanical parts in order to manifest the pattern. This, needless to say, was incredibly labour intensive, requiring close cooperation and synchronisation between the two operators. Weaving fabric in this way was slow and error prone.

Joseph Marie Jacquard, inspired by the work of Bastile Bouchon before him, made the first successful change to this in 1804. His pioneering work was a mechanism that fed cards through various spring rods, each of which controlled one of the mechanical parts formerly operated by the draw boy. These cards would trigger the springs by having holes cut out in them, to describe the pattern of the cloth, removing one operator and allowing the machine to perform much of the repetitive work. This was one of the first examples of machine automation, and one of the harbingers of the industrial revolution.

Such was the impact of the Jacquard loom, part of the history is written into popular culture today. As the Jacquard loom made it over to England, it deeply impacted the textile industry to the point where the industry protested its uptake, for fear of losing jobs. The movement, started in Nottingham, became known as the Luddites, and were famous for breaking the machines through various methods. This got so bad that Parliament eventually passed the Frame Breaking Act 1812, to render the breaking of these machines punishable by death. This is a pattern that plays out often in the world of technology, often signalling a paradigm shift in an industry.

What nobody realised at the time, however, was how significant an invention the punched card was. Punched cards would rule the roost in the world of computers for nearly a century and a half, and can still be found in use today in some very old organisations.

Daughter of a Poet, and the Father of the Computer

An alumnus of Trinity College, at the University of Cambridge, Charles Babbage was frustrated by the difficulty at producing trigonometric and logarithmic tables, a process that often took a very long time, and was incredibly error prone and repetitive. These three properties, in this combination, are something we've seen before, in the aforementioned loom. This is a recurrance we shall see a lot in this history.

The result was a machine, announced in 1822 to the Royal Astronomical Society, known as the Difference Engine. The machine could compute polynomials mechanically, working on a combination of gears and shafts, all rotating at various speeds in step with one another. While this isn't the first example of mechanical computation, it is the first example of "generic" mechanical computation with any known complexity - previous attempts were either for specific purposes, such as astronomical calendars, or incredibly simple, such as an abacus or slide rule.

Due to limitations on the technology at the time, the Difference Engine could never be built. A scaled down model was built, costing more money than was intended for the full engine, which was later demonstrated to Lady Byron, former spouse to George Gordon Byron, 6th Baron Byron, an English poet. Soon thereafter, work was suspended for cost reasons.

Undeterred, Babbage then went on to start development on his second machine. The Analytical Engine took the idea of the Differential Engine a step further, and was a machine that looks far more like a modern comptuer, from an architectural standpoint.

One of the crucial developments was the idea of an instruction sequence - the machine could take two numbers and add them, then multiply by a separate number from somewhere else, then go back and do the whole sequence again if it wasn't zero, for example. Therein was another key development - control flow. A key part of modern computing is "if this happens, do that". The analytical engine was the first machine to ever incorporate something like this. Drawing inspiration from the Jacquard loom, Babbage also incorporated sequences of punched cards into the machine, for instructions, and another, for the data.

Ada Lovelace, the daughter of Lord and Lady Byron, became deeply interested in Babbage's work, working closely with him as the Analytical Engine became more developed. Her most notable development was the construction of a sequence of instructions for generating Bernoulli numbers. This is widely considered to be the first computer program, making Lady Lovelace the first computer programmer.

Also visible at BeaR&D