Simon described the genesis of Feranti's Mark I and I* computers and their early history in the late 1940s and early 1950s. This was a seminal period in the history of computing.
I was just too young to experience it, but the first two computers I programmed were Ferranti Machines. Back in 1958, as an 11-year old, I wrote a short program in Pegasus Autocode; I finally debugged it a few years ago! Some years later, in 1966, I was lucky enough to get a gap year job working for David Hendry on the BCL compiler for Atlas, then the most powerful computer in the world.
I knew very little about the Ferranti Mark I which was the predecessor of the Pegasus. Like all electronic computers of those early days the Mark I used valves (vacuum tubes) which were prone to failure when their filaments broke.
Apparently the author of the Mark I matrix inversion routine was concerned that the software run-time (measured in hours) was close to the mean time to failure of the hardware and sought reassurance from Tom Kilburn, one of the machine's designers.
I can remember similar concerns about the risks of valve failure on Pegasus. Engineers would start the day by running the metal end of a screwdriver along the racks of hardware, They hoped that fragile valve filaments would break then and there, allowing the engineers to replace them before the day's work proper began.
One of the many remarkable revalations from Simon's talk was the composition of the Mark I programming team. Back in 1951 the majority of the programmers were women. One of them was Mary-Lee Woods. She managed to negotiate pay parity for the female programmers, a remarkable achievement for the time. Mary-Lee later married Conway Berners-Lee. You've probably heard of their son Tim!
Olaf Chedzoy, one of the early programmers from the Mark I programming team attended last night's meeting. He gave a brief but moving account of his work at Ferranti. He was responsible for the bootstrap code which read programs form 5-track paper tape into the Mark I*'s memory; his design was influenced by advice from Alan Turing, who worked on the loader for the Mark I.
Today's programmers work in an utterly different environment. The Mark I cost between £80,000 and £95,000. Today's equivalent would be about £3M. The Mark I had less memory, and much less processing power than a £12 Arduino.
Yesterday's talk left me with a thirst for more stories about the Ferranti Mark I and its creators. Fortunately there is a book on the way from Professor Lavington! It will be published later this year by Springer, and its provisional title is
Early computing in Britain: Ferranti Ltd. and government funding, 1948 – 1958.