The elusiveness of software: In the absence of hard models, rules of thumb have tended to govern a field in which it's more important for a program to be useful than to be totally correct

No theories of software reliability have been developed that are as powerful or as widely accepted as theories of hardware reliability. Some software specialists even question whether there is any practical meaning in such concepts as mean time to failure or mean time between failures; they argue that the distribution of software failures over time may have no mean value. Many workers therefore speak only of error rates, or of the estimated total bugs remaining. It is widely believed that even the most thoroughly debugged software still has glitches. The only vital requirement is that the program be useful, although not necessarily 100 percent correct.