The History of Language Processor Technology in IBM

Introduction In 1953 IBM introduced an early “automatic-programming” system: Speedcode for the IBM 701 computer. The goal of the system was to [l] “. . . lessen the enormous burdens of the programmer by providing a larger and more convenient instruction repertoire than a given machine provides.” In the same paper John Backus and Harlan Herrick go on to state: “There are two principal methods by which automatic-programming systems make these nonmachine operations available to the programmer: the interpretive method and the compiling method.” In the almost 30 years that have intervened since these observations were first made, methods for solving “the programming problem” have become more diverse and sophisticated, but the basic problem remains. Language processors-compilers, interpreters, macro systems, and assemblers-are still the principal methods used. In this paper we trace the history of IBM’s contributions to the techniques used in today’s language processors. The paper concentrates on IBM’s scientific and technological contributions, covering only widely used, general purpose techniques or those of particular scientific or historical interest. ment into its component parts were both elaborate and language-specific. The problem became the subject of numerous theoretical investigations and spawned a subfield of computer science: formal language theory. Today, elegant, language-independent parsing systems exist and are a common component of production compilers. The synergistic relationship between theory and practice continues to play a vital role in the development of language processor technology. This relationship is a subtheme of the paper: Particular attention is paid to tracing the origins and evolution of various techniques.