Response from R. A. DeMillo, R. J. Lipton, A. J. Perlis

the text is written in sometimes very poor English-"It is exactly those processes which mediate (sic) proofs of theorems in mathematics that require (sic) that..."-and their argtmlents are rambling. Supposing that they had something sensible to saywe can only regret that they have buried it under so much insinuating verbiage. As it stands it leaves the reader wondering why they have put so much venom in their text, because they seem to have gone much farther than the usual practitioner's backlash. None of the manypapers about program verification and derivation that I have written or seen uses APL as a programming vehicle. This could be an accident, it could also be a consequence of the rich expression structure of APL. (They refer to proofs as "...substitutions to be checked with the aid of simple algebraic identities" which, in the case of APL-expressions are perhaps not so simple...) If the latter conjecture is correct, it would explain why APL-addicts might feel unhappy about (or threatened by) modern achievements in proving the correctness of (non-APL) programs. Does it help the understanding of this paper and its venom to know of the heavy involvement with APL among its authors? We can only guess and have our private opinions. We are grateful to the editor of SEN for obtaining Professor Dijkstra's permission to publish [the above memorandt~n] and for giving us the opportunity to respond to it. We must begin by refusing to concede that our confidence in a piece of real software has ever been increased by a proof of its correctness-real software, such as sirline ticketing programs that issue real airline tickets or classroom scheduling programs that schedule real classes. We appreciate that Professor Dijkstra's memo was quite informal and was originally intended for those sympathetic to program verification, and among them it may be true that cases of raised confidence are well known and require no elaboration. But now that the correspondence has reached a larger audience, we hope that Professor Dijkstra will be quick to substantiate his assertion that such cases exist by pointing to some real-life examples. In the form in which it is now expressed, the question about confidence may lead to an impasse: "Doesn't increase our confidence." "Does mine." "Doesn't." "Does." Our best recourse may be to the general practice of our peers. Do they act as if prograu verifications increased their confidence in programs? Do adherents …