Saying What You Mean

Specifying a problem and its solution is often viewed as dreary work. However, identifying the essential properties of a problem, and stating them clearly, is the heart of problem solving. Careful consideration of what a problem requires often leads to a solution, and clear and precise specification is necessary to determine whether a solution is correct. Formal specification is not only the necessary basis for any formal verification, but much of the work--and the most interesting work--of verification is in writing a specification, and iteratively adjusting it until it is suitable. In this talk, I will discuss the value of good specification, both formal and informal, some of its often overlooked benefits, and challenges in doing it well (and indeed, what constitutes doing it well). I will use anecdotes, drawn mostly from my experience in developing and verifying concurrent algorithms, to illustrate both positive and negative examples.