The following is the text of Pathfinder, an article about John Backus published on pp. 18-24 of the July-August 1979 edition of Think, the IBM employee magazine.
They were an eager, absorbed young group that spring of 1954: three IBM computer programmers and a former U.S. Foreign Service employee hired to do technical typing. Their offices were tucked away on the 19th floor of the annex to what was then known as IBM World Headquarters — down the block from Tiffany’s on the busy corner of Manhattan’s 57th Street and Madison Avenue. Far below, in the ground-floor display center, was the machine they were trying to improve upon. It was the IBM 701 computer, which, only the year before, had launched the company into a brand new world of electronic data processing.
By November, they were ready with a preliminary report. Based on what the group’s manager, John Backus, then 29, calls more faith than knowledge.” It stated that the programming language they had designed for the new 704 would enable it “to accept a concise formulation of a problem in terms of mathematical notation and to produce automatically a high-speed 704 program” for its solution. The report suggested that the automatic program would run as fast as a program painstakingly coded by a human programmer. Months of testing would prove them right. They named the language FORTRAN, for FORmula TRANslation.
Backus, now [in 1979] an IBM Fellow, went on to become a staff member of the Thomas J. Watson Research Center in Yorktown Heights, N.Y. Sixteen years ago, he traded the East Coast for California and the IBM Research Laboratory in San Jose. Of the group’s other members: Irving Ziller is now a special consultant to IBM Vice President and Data Systems Division President John E. Bertram; Harlan Herrick is a systems engineer for the General Systems Division in Manhattan; the typist, Robert A. Nelson, who showed a talent for FORTRAN design, is an IBM Fellow and an important contributor to the virtual storage concept. As for the 704, IBM’s first machine with magnetic core memory has long since become a museum piece. And last year, to make way for IBM’s new 43-story skyscraper now under construction, the old headquarters annex fell to the wrecker’s jackhammer.
But though the FORTRAN primer is more than 20 years old, and subsequent languages have sprung forth like dandelions on an April lawn — some 165, at last count — FORTRAN has proved exceedingly durable. Adapted to subsequent computer models, it is, today, one of the most widely used computer languages in the world. Three years ago, Backus went to the White House to receive the nation’s top award for scientific and engineering achievement, the National Medal of Science, for pioneering contributions to computer programming languages.
These days, Backus divides his time between the San Jose lab and his cliffside home near San Francisco’s twin peaks, where he lives with his writer wife, Una Stannard, and maintains an office with a spectacular view. He answered his front doorbell one afternoon not long ago, and led me up a flight of stairs into a bright living area filled with wicker furniture, a variety of healthy green plants and abstract paintings. As we seated ourselves by the window, the city, far below, gleamed white in the afternoon sun.
Now that FORTRAN is programming history, I wondered what its author was doing for an encore. Backus jumped up quickly and disappeared into the next room, returning with a magazine reprint in his hands. “I’ve had an article published in Communications,” he said, modestly neglecting to mention that it was the Turing Award address he had given before the Association for Computing Machinery, in recognition of his early technical work. “The import of what I said,” he continued, “is that conventional programming languages, including FORTRAN, are very poor languages for telling computers what to do, basically because you can’t say very much. The kinds of languages I’m into now are radically different. It’s all very exciting for me, because this is what it was like back in the FORTRAN days. I mean, it’s this completely new field, with all kinds of questions coming up.”
For a “pioneer,” Backus is a young 54. He wears jeans as lithely as a teenager. When he talks, he gestures emphatically.
“I guess the best analogy comes from the development of mathematics,” he explained. “Mathematics, you know, started with arithmetic, and then it got into slight abstractions, like simple algebra, simple equations, and then it got into questions of the structure of algebraic laws for the operations of arithmetic. What we’ve been stuck with in programming is analogous to the arithmetic stage. And what I’m trying to do is move from that hideously complicated manipulation of numbers up to abstractions, where you have structure and you can reduce a whole set of rules to one simple rule. If I succeed, hopefully, we’ll have an intellectual foundation for a lot of new computer designs.”
Surprisingly enough, for one who has his master’s degree in mathematics, Backus didn’t set out to be a mathematician. The son of a Wilmington, Delaware, chemist-turned-stockbroker, he had, in his words, a “checkered educational career.” In and out of prep school from the 8th grade on, he spent six months at the University of Virginia, marking time until the Army draft. He had thought he might become a doctor and, once in the Army, he studied premed and began medical school at what is now New York Medical College. “I had visions,” he recalls, “of right away doing research on the functions of the brain. But at medical school, all they wanted you to do was memorize, memorize, memorize. By the time I got out of the Army, my one ambition was to build a good hi-fi set.”
It was at radio technician school, under the G.l. Bill, that Backus discovered math. He went on for his master’s at Columbia University and, while a student, went down one day to Madison Avenue and 57th Street to have a look at what a classmate had described as “an interesting thing.”
The interesting thing, it turned out, was the IBM Selective Sequence Electronic Calculator (SSEC). Installed at IBM headquarters in 1948, it was the response of company engineers to IBM’s belief that electronics was the new growth area. In the early postwar years, the first electronic computer, the ENIAC, was nearing completion at the University of Pennsylvania’s Moore School of Electrical Engineering; the mathematical genius of John von Neumann was bringing the stored program concept to realization at Princeton’s Institute for Advanced Study; while at Harvard, IBM’s own Mark I Automatic Sequence Controlled Calculator was doing multiplication and division in seconds.
Equipped with 13,000 vacuum tubes, 23,000 relays and a large number of paper tapes, the SSEC was a hundred times faster than the Mark I. The company that supplied its lubricants advertised it in The Saturday Evening Post as the Oracle on 57th Street. Customers used it to design turbine buckets and solve oil field exploration problems. And Wallace J. Eckert, the director of IBM’s Watson Scientific Computing Laboratory, did SSEC calculations of the moon’s orbit that would show up 20 years later in the Apollo space program.
Heady stuff for a young math major. When the IBM systems service rep who put the SSEC through its paces learned that Backus would soon be in the job market and suggested he talk with the machine’s co-inventor, Rex Seeber, he offered little resistance. “I had holes in the sleeves of my jacket and my shoes needed shining, but she got me an interview then and there. Seeber gave me a little homemade test and hired me on the spot.”
The next two years, spent computing lunar positions, were “just delightful. You had the machine for two weeks all to yourself, just to check out your tapes and plug-boards and things like that. And then, of course, you had to be there the entire time the program was running, because it would stop every three minutes, and only the people who had programmed it could see how to get it running again.”