EN
Back to the archive

The encyclopedia · Software & IT · Technical decision · 1953–1966

FORTRAN let engineers write formulas directly and became the first language standard

John Backus's 1957 compiler turned math-like formulas into code nearly as fast as hand-written machine language, so scientists stopped hiring programmers.

IBM

The solution

In the early 1950s programming meant writing machine code by hand. John Backus called it "hand-to-hand combat with the machine": programmers spent up to half their time debugging, and coding labor cost at least as much as the computers themselves.

In 1953 Backus got IBM funding for a small team to test a better way. FORTRAN (formula translation) was created in 1954 and released commercially in 1957 on the IBM 704. It let scientists and engineers input problems in something close to mathematical notation: what once took about a thousand hand-keyed instructions could be expressed in 47 FORTRAN statements.

Skeptics said compiled code could never match hand-crafted code. Backus's team built the first optimizing compiler, producing code that ran nearly as fast as hand-written numeric code; and because other vendors adopted IBM's standard, a FORTRAN program could run on any machine with a FORTRAN compiler.

Why it worked

  • Formula notation matched how engineers already thought.
  • The optimizing compiler removed the speed penalty that blocked adoption.
  • IBM let other vendors implement the same standard, so the language became portable.
  • It cut the cost and labor of coding dramatically.
What it achievedLet users write formulas; the compiler does the translating.clever

What can be applied

A tool wins by matching the user's existing notation, then making the new path as fast as the old one; portability follows when the standard is open to every vendor.

Aftermath

By the mid-1960s FORTRAN was the first national computing standard in the US and ran in most major data centers. It is still used in weather forecasting, physics simulations and genomics; Backus later won the Turing Award (1977) and the National Medal of Science (1975).

Sources

spotted an error? The archive wants to know.

Related cases