Since the question "Does program P obey specification S" is undecidable in general, every practical software validation technique must compromise accuracy in some way. Testing techniques admit the possibility that a fault will go undetected, as the price for quitting after a finite number of test cases. Formal verification admits the possibility that a proof will not be found for a valid assertion, as the price for quitting after a finite amount of proof effort. No technique so dominates others that a wise validation strategy consists of applying that technique alone; rather, effective validation requires applying several techniques.
