I first used Fortran in the autumn of 1963, on a Univac 1107. The language was known as Fortran IV and was broadly comparable in facility to what was implemented on the IBM 7090/4 series; the hardware similarly had 36-bit words and 6-bit characters. The other main languages in use at the site were Cobol, Algol 60 and Simula I, plus the assembly language SLEUTH (Symbolic LanguagE for the Univac THin film computer).

Fortran was the standard language at the site for scientific and technical work. I worked on various projects including ones to do with finance, civil engineering and environmental work, and also on building up the library of statistical programs. The latter included implementing a large library of programs from the US written in Fortran II. Even at that time Fortran II, with its intrinsic functions ending with 'F', certain statements having a 'D' in column 1 and with its 'sense switch' statements appeared quaintly old-fashioned. Fortunately there was a translation facility, SIFT, which did much of the basic clerical work.

The 1107 had 15 index registers and 16 accumulators, 5 of which overlapped. This allowed for some clever programming at machine-code level and the compiler-writers made full use of the possibilities. With hindsight it is remarkable that compiling, particularly optimization, techniques had developed so quickly in the relatively short life of high-level languages.

For me, coming from a site that had used Mercury Autocode with input on 5-track paper tape, it was liberating to be able to use Fortran with 80-column punched cards. In 1967 I went on to use a Univac 1108, which had an extended language called Fortran V, before moving in 1970 into the IBM 360 world with its choice of compilers, Fortran G, Fortran H and WATFOR/WATFIV.

Next came the ICL System 4. Working at a university that had its own locally written operating system, with its own locally written language compilers, meant that none of the applications software normally expected to be available in an academic environment could be acquired and installed in the normal way. Everything, from a variety of original platforms, had to re-implemented from source code, which was a hugely time-consuming process. It was not all wasted effort however; it became normal to find quite large numbers of errors even in well-established and widely used software. Apart from use of non-standard statements, these were typically caused by use of uninitialized variables and by subscripts going out of range.

It was this experience which led me to become interested in language standardization.

David Muxworthy
BSI Fortran Convenor
November 2006

---

Comments on this or any other of the Group's pages should be sent by email to the FSG Web Editor.

Last modified: Sun 9 Dec 2012 15:00:31