Abstract This paper proposes some improvements to the archaic methodologies currently used for checkout of real‐time software. The emphasis is placed on removing errors from the computer program source code in the host computer environment, and deferring target computer checkout until those errors have been largely removed. Specific tools and techniques to make this possible are described. This improvement is seen to be risk‐free and to reduce cost. It is, however, controversial.

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.