Fortran born of frustration

ALSO    Flashback: The history of computing index    Sign up for the Computer Connection email service    For more computing stories MORE FROM COMPUTERWORLD.COM  • Computerworld's home page  • Flashback timeline  • Computerworld Year 2000 resource center  • Computerworld's online subscription center REVIEWS & IN-DEPTH INFO AT IDG.net  IDG.net home page  • IDG.net's personal news page  • Search IDG.net in 12 languages  • Subscribe to IDG.net's free daily newsletter for IT leaders NEWS RADIO  • Computerworld Minute  • Fusion audio primers THIS YEAR IN COMPUTER HISTORY Technology Advances • An IBM/American Airlines joint research team releases a study that becomes the genesis for the Sabre computerized airline reservation system. • IBM sells 450 of its 650 magnetic drum calculator, outpacing the 701 Defense Calculator and establishing itself as the first mass producer of computers. • William Shockley, a Nobel Prize winner for developing the transistor, founds Shockley Semiconductor in Palo Alto, Calif. The company brings together the Fairchild Eight, the engineers who would go on to create Fairchild Semiconductor and, indirectly, Silicon Valley. Born in 1954 • Susan Kare, legendary Macintosh designer (smiling computer, wristwatch, trash can) • Scott McNealy, CEO of Sun Microsystems • Oprah Winfrey • Philip Zimmerman, creator of Pretty Good Privacy Other Notables • Best Picture: On the Waterfront • Number of homes in the U.S. with TVs: 29 million • First TV dinner is introduced: It costs 98 cents and contains turkey, dressing, gravy, peas and sweet potatoes. • Alan Turing, mathematician and artificial intelligence pioneer, commits suicide after being persecuted, and prosecuted, for homosexual acts.

Inventor of first popular computer language wanted to avoid tedium of machine language programming.

May 3, 1999
Web posted at: 3:55 p.m. EDT (1955 GMT)

by Leslie Goff

(IDG) -- Sometimes drudgery, rather than necessity, is the mother of invention.

John Backus says that's what inspired him to propose developing the first automatic programming language, Fortran, or formula translation. Then 29, the mathematician from Columbia University had spent several years working on IBM's 701 and 704 computers and was simply tired of the complexity of programming.

"It was pure laziness," Backus, now 74, says of his inspiration. "Writing programs was a big drag -- you had to have enormous detail and deal with things you shouldn't have to. So I wanted to make it easier."

Backus had to wing it all the way: There were no studies to support his ideas, no methodologies for developing software, no models for success.

In late 1953, Backus wrote a memo asserting that at least half to three-quarters of the operating costs of a computer were from programming and testing. "The salaries of programmers generally equaled or exceeded the rental [cost] of a computer," Backus says. He reasoned that if he could put together a team that could come up with a language to automate instruction code, it would bring costs down significantly by enabling a computer to perform a repetitive task from a single set of instructions by using loops.

Backus' idea wasn't original. Grace Hopper of Remington Rand's Eckert Mauchly division had created the A-O compiler, designed to do roughly the same thing. But the compiler "was clumsy and ran slowly and was difficult to use," Backus says. Moreover, IBM's new 704 computer incorporated two new features that would aid Backus' mission -- and make it more complex.

Built-in floating-point co-processing and index registers would let the 704 use automatic mathematical statements, which would eliminate the need to write repetitive instruction code. The improvements meant that the programming techniques had to be a lot more clever, Backus says, because "you couldn't mask inefficiencies."

The complexity, coupled with the lackluster response to the A-O compiler's performance, made for skepticism within and outside IBM. But Backus' boss, Cuthbert Hurd, then director of IBM's applied science division, gave him the go-ahead, and in early 1954, the work began.

Backus initially recruited Irving Ziller from within IBM to work with him and later added another IBMer, Harlan Herrick. He says he sought out those with "creativity, a lot of smarts and experience. We had a great variety of people: a physicist, a crystallographer, an English major."

Sheldon Best, on loan from MIT, wrote a difficult part of the program, Backus says: figuring out how to use index registers. "He would do a flow chart that started out on a piece of paper, and as he would add to it, he just kept gluing pieces of paper together into this whole enormous flow chart," Backus says. "When he went back to MIT, it took months to figure out what it all meant and how it worked."

Roy Nutt, head of the data center at United Aircraft -- and later founder of Computer Sciences Corp. -- was so enthusiastic about the language that he also joined the gang. "Roy was known to just sit down at a keypunch machine and keypunch in a program that would run. He was responsible for the whole input-output system in Fortran," Backus recalls.

The programmers put in late hours -- sometimes sleeping during the day at the Hotel Langdon, across from the IBM building on New York's Madison Avenue, so they could secure computer time at night. Over two years, the team grew to include some of the best programmers to pass through IBM, Backus says.

The summer of 1956 was spent testing the language. IBM shipped the first copies of Fortran to customers in 1957.

Today, 42 years after shipping, Fortran remains a dominant language for military and scientific applications.

Goff is a frequent contributor to Computerworld. She can be reached at lgoff@ix.netcom.com.