Transition network grammars for natural language analysis
Communications of the ACM
A verified program-verifier.
A synthesizer of inductive assertions
AFIPS '75 Proceedings of the May 19-22, 1975, national computer conference and exposition
| Hi-index | 0.00 |
"A proof of correctness guarantees that a program will run correctly every time it is executed." That statement is not necessarily true. Suppose, for sake of concreteness, that a valid proof of a Fortran program has been constructed. When this program was proved, it most likely was proved in isolation from the other software components which ultimately will be involved in actually making the program run. So, even though we have proved the Fortran program, one of these other components, or the system hardware, may malfunction causing the actual machine language program that is executed to produce an error. These comments are not an argument against proving programs at the Fortran level, but rather an indication of the eventual need for a completely proved computing system.