PERSON
John Backus
He joined IBM in 1950, worked on the SSEC and Speedcoding, and from 1954 led the team that built FORTRAN — an optimizing compiler good enough that assembly programmers accepted its output, and the work for which he received the 1977 Turing Award. The formal notation he devised to describe the syntax of ALGOL 58 was reworked by Peter Naur for ALGOL 60 and, at Donald Knuth's suggestion, became known as Backus-Naur form. In later years he moved to function-level programming, presenting FP in his Turing Award lecture. IBM Fellow 1963, National Medal of Science 1975, Draper Prize 1993.
Profile
- Born
- 1924
- Died
- 2007
- Span
- 83 years
- Appearances
- 02
- Name
- ENJohn BackusJAジョン・バッカス
John Warner Backus (3 December 1924 – 17 March 2007) spent his career trying to strip out of programming the part that exists only to suit the machine. He succeeded twice in ways that changed the industry: once with a compiler, FORTRAN, and once with a notation, the syntax formalism of ALGOL 60.
Into IBM
Born in Philadelphia, he took an MS in mathematics at Columbia and joined IBM in 1950. His first three years went to the Selective Sequence Electronic Calculator, where his initial assignment was computing positions of the Moon. In 1953 he produced Speedcoding, the first high-level language written for an IBM machine.
He later described what programming was like before all this as "hand-to-hand combat with the machine" — and, as IBM's own account of Fortran adds, the machine usually won.
FORTRAN
In 1953 Backus argued that it should be possible to build a language that captured the human intent of a program and recast it into something a computer could process, and was given a budget to hire a small team to test the idea. Serious development ran from 1954; the first compiler for the IBM 704 shipped to customers in April 1957.
The decisive achievement was not the language design but the optimization. Working programmers of the period believed high-level code was necessarily slow, and they were right about the code that existed. FORTRAN broke that assumption: only once a compiler emitted machine code that assembly programmers would accept did compiled languages become a real option. The 1977 Turing Award was given for this work.
A notation for syntax
After the 1958 Zurich conference produced the preliminary report on ALGOL 58 (IAL), Backus built a formal notation to describe its syntax precisely and presented it at the ICIP conference in Paris in June 1959. The ALGOL 60 report adopted the notation and cites that paper in a footnote. It was the report's editor, Peter Naur, who reworked it for ALGOL 60.
The name settled in 1964, when Donald Knuth published a letter arguing that since "Backus Normal Form" is not a normal form in any sense, it should be called Backus Naur Form — which also kept the familiar abbreviation and credited both men.
The practice of writing a language's syntax as a set of rules rather than as prose starts here and has never stopped. Parser generators, and the grammar sections bolted to the back of protocol specifications, are all downstream of it.
Leaving the von Neumann style
Backus did not stop there. His 1977 Turing Award lecture was titled "Can Programming Be Liberated from the von Neumann Style?", and what it presented was FP, a system of function-level programming. His argument was that the habit of building programs out of assignments to variables is itself what makes them hard — an objection, from the inside, to the very style FORTRAN had helped make universal.
FP never became mainstream. But the seriousness the functional-programming argument acquired during the 1980s owes a good deal to that lecture.
Honours
- IBM Fellow, 1963
- IEEE W. W. McDowell Award, 1967
- National Medal of Science, 1975
- ACM Turing Award, 1977
- Fellow of the American Academy of Arts and Sciences, 1985
- Charles Stark Draper Prize, 1993
He retired from IBM in 1991 and died in Ashland, Oregon, on 17 March 2007.
Appearances
Last updated: