A Colossal ESC-Brazil 2013

I'm recently returned from the Embedded Systems Conference in Brazil. It was a short, two day event that packed in many hundreds of visitors. The show floor was surprisingly enormous; I estimate it to be five times larger than the last time I attended this event two years ago.

That show floor is very different from what regular attendees to the US shows might expect. The usual suspects are there: ARM, Agilent, Tektronix, LDRA, and other suppliers to the embedded space. But there was an array of other companies, like those who provide machine tools, medical equipment, plastic-forming machines, and more. Not speaking Portuguese the logic behind such a wide array of vendors escapes me, but those non-embedded booths found themselves swarmed by visitors.

Some familiar faces presented talks, like Bill Gatliff, Chris Shore and Steve Mellor. Bill questioned the headlong rush to include Android in, well, everything. Chris gave a lot of insight into doing battery-powered design, and, alas, I missed Steve's talks due to schedule conflicts.

Chris also gave an hour-long presentation on the history of the Colossus, possibly the most seminal of all early computers. It's a story that was long buried; used to help decrypt German ciphers during WWII the British government didn't officially recognize the machine's existence till the year 2000. Earliest hints about this 1944 machine, or series of machines (a dozen were built) started to surface in the `70s, some three decades after its role in shortening the war.

Though ENIAC's patents were tossed out by a court in the `60s it has long held the honor of being the "first" electronic digital computer. But that honor was muddied by the earlier Colossus's secrets. With 18,000 vacuum tubes, ENIAC was a marvel of engineering. But it was rather like a huge calculator at first, designed merely to compute ballistics tables.

The story of Germany's Enigma encryption machines is old and often told. Less well know are the Lorentz devices, which mashed a pseudo-random stream of binary 22 million words long against the plaintext. Chris showed that the scientists on the trail of these codes found that the Lorentz messages were very long - thousands of characters, and largely originated from Berlin. The boffins at Bletchley Part (England's decryption center during the war) realized these messages contained detailed orders from the German high command. Their value was priceless.

The code was very tough to break. Patterns were discovered, but a single message took five weeks to decrypt, by which time the intelligence was obsolete.

Eventually (and leaving out a lot of the fascinating story) they built the Heath Robinson, a Rube Goldberg-like contraption that played two punched tapes against each other, looking for particular patterns of data. The manual five-week process was reduced to 18 minutes. On a good day, because the tapes stretched, broke, and the machines themselves were unreliable.

Enter Tommy Flowers, a self-taught EE who realized vacuum tubes (or, "valves" to our friends across the pond) could improve the process. He proposed a machine with an astonishing 1500 tubes, growing to 2500 in later versions, where one of the paper tapes was replaced by circuitry.

Flowers invented this machine from thin air. The work on ENIAC was secret so he was unaware of it. Chris said this was the first use of an electronic shift register in a computer; the first use of a single clock; possibly the first use of interrupts (to manage the printer). The optical tape reader ran at an unprecedented 10k characters per second (though was usually throttled to 5000). I was struck by that, as 30 years later we were using tape readers for microprocessor development that maxed out at 200 CPS.

What intelligence did Colossus produce? After all, it didn't come on line until early 1944, just over a year before the war in Europe ended. Chris told of how the D-Day invasion had been postponed, postponed, and put of yet again. On June 5 Eisenhower and his generals were meeting about launching the invasion the following day. No resolution could be met; if they delayed, that would cost a month due to the tides. Suddenly an aide rushed in with a Colossus decrypt. Hitler had sent Rommel a message indicating he was convinced there would be a faint at Normandy and the real invasion would come at Calais days later. He ordered Rommel to keep his tanks and heavy armor in position there. Eisenhower read the message and announced the invasion would take place the next day.

Over the years the ESC has had some great keynote speakers. Douglass Adams - one of the best. Robert Ballard - wow!

And this keynote by Chris Shore: possibly the best computer-related talk I have ever attended.

Published August 28, 2013