Société d'électronique et d'automatisme

~ the eulogy ~

François-Henri Raymond was sent to the United States in 1947 by his employer, a maker of radar and television equipment, with no particular brief. He met Howard Aiken at Harvard and came away with John von Neumann's report on the EDVAC, which he is said to have bought in Washington for a few dollars. Back in Paris he told his employer that France should build these machines. His employer was not interested, so he resigned, and in December 1947 he started the Société d'électronique et d'automatisme in a hall of a former Delage automobile factory at Courbevoie that had been bombed during the war. The company was legally constituted the following March. Its first customer was the Air Force's missile bureau, which needed arithmetic done.

The analog machines came first and paid the bills longest. The OME 11 appeared in February 1949, and the series that followed sold close to two hundred units with a real export business behind it. They were used for nuclear and hydrodynamic simulation, and later for flight simulations of Concorde. The digital work was harder. A contract with the armaments directorate in 1951 produced CUBA, probably the first stored-program computer built in France, and almost every component decision on it ran ahead of what French industry could supply. The magnetic core memory had no French manufacturer, and the drum had to be bought from Ferranti in Britain. SEA was the first French company to use ribbon cable and wire wrap. CUBA was delivered in 1954 and running by 1956, by which point it was obsolete and was abandoned shortly after.

What worked commercially was smaller. The CAB 1011 went into service in July 1955 and ran until 1965 doing cryptanalysis for the French foreign intelligence service. The CAB 2000 and 3000 families followed as the company's first machines built in series. The French word for a computer, ordinateur, was not coined until 1955, which is a fair marker of how early all this was. At its height the company employed up to 800 people and held close to a thousand patents, and it installed around 170 computers between 1955 and the end.

~ the dossier ~

~ the desk machine ~

The CAB 500 is the one that should be better known. From 1956 SEA invented a logic gate called the Symmag, built from ferrite beads like those in core memory, slower than transistors but reliable enough to make a small machine practical. Around that it built a computer that sat on a desk and was meant to be operated by one person with no technical training, driven by an interactive high-level language called PAF. The first unit went to a customer in February 1961, sixty were in service by September 1962, and about a hundred were built, with a dozen or so exported as far as China and Japan. Alice Recoque, later a significant figure in French computing, worked on its development.

The CNAM calls SEA the first computing startup in France if not in Europe, and describes the CAB 500 as arguably the first personal computer, which is a claim worth handling carefully but not worth dismissing. The machine was a decade and a half ahead of the Altair and aimed at the thing the Altair eventually delivered, which was one person sitting down at a computer without a specialist standing between them.

~ plan calcul ~

Everything after this point is politics. On 20 March 1963 the government pressed SEA into an agreement with Bull: Bull would sell SEA's range, and SEA would get Bull's factory at Les Andelys. It went badly in both directions. Bull's salesmen had no idea how to sell to scientific customers, so the CAB 500 underperformed, while Bull found itself carrying the CAB 3900 business machine against its own competing product. Then came the Bull affair, when General Electric took control of the French champion, and SEA lost further ground. A 1964 understanding with Control Data went nowhere once CDC set up its own French subsidiary.

SEA's answer was ambitious and never shipped. Watching IBM announce the System/360 in 1964, it drafted a compatible family of its own built on stack-machine principles with Algol as the machine language, planned an entry-level CAB 1500 and a large CAB 15000, and contemplated manufacturing a thousand of them. In December 1966 the government, executing the Plan Calcul, compelled SEA to merge into the Compagnie européenne d'automatisme électronique, a joint venture of CGE and CSF roughly three times its size, to form the Compagnie internationale pour l'informatique. The prototypes were stopped, the teams were broken up over the following year, and Raymond was moved aside inside the Thomson group. At the end it held about 3.1 percent of the French market.

~ inventory 22358 ~

Raymond went to the Conservatoire national des arts et métiers in 1967 and taught computing there until 1979. He died in Toulouse on 25 November 2000. At the first colloquium on the history of computing in France, at Grenoble in March 1988, he presented a paper on his own company under the title "Une aventure qui se termine mal", an adventure that ends badly, and a second one on the Plan Calcul that put the blame where he thought it belonged. He described what was destroyed as the humus of twenty years of teamwork, wrecked stupidly and unjustly.

A CAB 500 survives in the Musée des Arts et Métiers under inventory number 22358. It took a researcher until 2004 to find it, in the museum's reserves with its drum disconnected and stored alongside, after three institutions had each declined to confirm they held the thing. Permission to restore it was refused. The company that trained the first generation of French computer engineers ended as a clause in somebody else's merger, and its best machine is an inventory number in a store room.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »