When I signed up as a trainee programmer, in 1965, I was expecting to work on the LEO II - the successor to the world's first commercial mainframe computer.
What I worked on was System 4 - a copy of an IBM computer. "Why?" I asked. "Is the IBM design better?"
No, said my boss. "If we have a machine that will run IBM's software, then there's a lot of software that people will want to run on our machine because ours is cheaper. And IBM sells a lot more computers than we do."
What a week this is. Twenty years since the invention of the commercial IBM PC; 50 since the invention, in Britain, of the commercial mainframe computer. So why doesn't our computer technology dominate the world?
To understand, you have to stop asking "Why didn't LEO become a success?" It was a huge success. But it wasn't what the future wanted - a range of compatible machines.
Before IBM invented the System/360 range, a small computer could not do the things a large computer could. With the /360, the small, medium and big machines were virtually identical in function and capability; the variation was in the workload. A program written for a small /360 would run on a large /360, making it possible for an enterprise to start small and grow.
The sensible question would be: "Why didn't Lyons Electronic Office turn LEO into a range?" The man who headed up the LEO project, David Caminer, saw the problem as one of lack of resources. Lyons was a chain of tea shops, not a computer manufacturer. Having sold a dozen LEO II computers and more than a hundred LEO III models, it sold out to English Electric, which abandoned LEO technology.
But there was another problem: a failure of vision in the industry - at Ferranti, GEC, Plessey and even LEO's successor, ICL. Governments of the day saw military computing as the focus
Today, we measure computers in terms of how many millions of transistors on a single processor chip. But LEO didnot have a transistor, because it predated its invention. Its successor, LEO II, was a different machine. The same major break was made with the transistorised LEO III - which was the workhorse of the bureau where I worked.
Lyons wasn't pinning its hopes on LEO III, because it couldn't. When I programmed LEO III, it was too small to run Cobol, the language of the day. We wrote our work in Cleo - a cut-down Cobol-like language. When we looked to the future, we wrote System 4 Usercode.
System 4 was the great hope of a consortium of designs. It was to be built by a new computer company formed by English Electric, LEO and Marconi, all of whom decided IBM was right. They were ahead of the field because they wanted a range that ran the same software as IBM machines. And RCA had designed the Spectra, which did exactly that. So EELM licensed the Spectra, and called it System 4; and proposed to sell it to the Post Office, to handle the Giro.
The dead hand of Government sponsorship was immediately felt. It was considered desirable to make EELM a big ger company to supply the Post Office - which was holding out for an IBM /360 contract.
So on to EELM, Frankenstein's monster. Full of incompatible dreams and designs, they bolted on the best-selling ICT - International Computers and Tabulators. The result was called ICL, International Computers Limited. It was a mess, but it got the Giro contract.
British industry needed to set up a credible rival to IBM. The giants of computer design were diametrically opposed to the IBM range concept. They were doing unique designs, mostly for defence contracts, working on the principle that the correct way to design a computer, was to examine the task for which it was going to be built. Then you designed a programming language that best expressed the solution. And then, you designed the optimum machine to run that language.
ICL carried on selling its ICT-derived 1900 range. It wasn't a true range, but it had some of the characteristics, and software written for a small 1900 could be transported to some large ones. It designed the new range in the early 1970s; a decade too late. By that time the world was hooked on IBM /360 code.
Nobody in ICL tried to penetrate the American market. ICL's 2900 struggled to match IBM's /370 range, and finally, the managers gave up, and sold out to Fujitsu.
This year Fujitsu gave up trying to keep the name ICL alive, and the last traces of LEO disappeared.
