Information Processing Requirements for Future Air Force Command and Control Systems, and Some Implications for Software Research and Development.
Summarizes the complex tasks confronting computerized command and control systems, and describes the lack of related quantitative information about computer programming and usage. Six of the few research studies are cited, showing that widely held ideas are invalid: (1) There can be a significant software learning curve; three successive FORTRAN compilers required 72, 36, and 14 man-months. (2) Most program instructions are simple; of 100 typical FORTRAN statements, 68 are A=B and 24 are A=B+C. (3) Software checkout and testing takes as long as design and development. (4) To minimize total hardware and software costs, overbuy computing capacity by 50 percent to 100 percent; cost per instruction mushrooms as computers approach full utilization. (5) In one experiment, reduced computer availability improved users' performance. (6) System designers tend to consider their own usage patterns without investigating users' needs, which are usually different. Computer "science" will not be soundly based until scientific methods are employed. (Presented at an ONR/SDC technical symposium.) 26 pp. Ref.