Key points are not available for this paper at this time.
Persistent software errors-those which are not discovered until late in development, such as when the software becomes operational-are by far the most expensive kind of error. Via analysis of software problem reports, it is discovered that the predominant number of persistent errors in large-scale software efforts are errors of omitted logic..., that is, the code is not as complex as required by the problem to be solved. Peer design and code review, desk checking, and ultrarigorous testing may be the most helpful of the currently available technologies in attacking this problem. New and better methodologies are needed.
Robert L. Glass (Sun,) studied this question.