Standards

Remarkably, C achieved its success in the absence of a formal standard. Even more remarkable is that during this period of increasingly widespread use, there has never been any serious divergence of C into the number of dialects that has been the bane of, for example, BASIC. In fact, this is not so surprising. There has always been a “language reference manual”, the widely-known book written by Brian Kernighan and Dennis Ritchie, usually referred to as simply “K&R”.

The C Programming Language,
B.W. Kernighan and D. M. Ritchie,
Prentice-Hall
Englewood Cliffs,
New Jersey,
1978

Further acting as a rigorous check on the expansion into numerous dialects, on UNIX systems there was only ever really one compiler for C; the so-called “Portable C Compiler”, originally written by Steve Johnson. This acted as a reference implementation for C—if the K&R reference was a bit obscure then the behaviour of the UNIX compiler was taken as the definition of the language.

Despite this almost ideal situation (a reference manual and a reference implementation are extremely good ways of achieving stability at a very low cost), the increasing number of alternative implementations of C to be found in the PC world did begin to threaten the stability of the language.

The X3J11 committee of the American National Standards Institute started work in the early 1980's to produce a formal standard for C. The committee took as its reference the K&R definition and began its lengthy and painstaking work. The job was to try to eliminate ambiguities, to define the undefined, to fix the most annoying deficiencies of the language and to preserve the spirit of C—all this as well as providing as much compatibility with existing practice as was possible. Fortunately, nearly all of the developers of the competing versions of C were represented on the committee, which in itself acted as a strong force for convergence right from the beginning.

Development of the Standard took a long time, as standards often do. Much of the work is not just technical, although that is a very time-consuming part of the job, but also procedural. It's easy to underrate the procedural aspects of standards work, as if it somehow dilutes the purity of the technical work, but in fact it is equally important. A standard that has no agreement or consensus in the industry is unlikely to be widely adopted and could be useless or even damaging. The painstaking work of obtaining consensus among committee members is critical to the success of a practical standard, even if at times it means compromising on technical “perfection”, whatever that might be. It is a democratic process, open to all, which occasionally results in aberrations just as much as can excessive indulgence by technical purists, and unfortunately the delivery date of the Standard was affected at the last moment by procedural, rather than technical issues. The technical work was completed by December 1988, but it took a further year to resolve procedural objections. Finally, approval to release the document as a formal American National Standard was given on December 7th, 1989.