Concurrent programming for the masses (invited address)

This talk is concerned with the question: If hundreds of thousands of people are going to be writing distr ibuted programs, what sort of programs should they write? This question raises several questions in turn: Are hundreds of thousands of people going to be writing distr ibuted programs? If research scientists find it difficult to write correct distr ibuted programs (and they do find it difficult, as evidenced by published incorrect dis tr ibuted algorithms), wouldn' t it be bet ter to steer ordinary mortals away from distr ibuted programs? Will distr ibuted hardware be widely available? A different kind of question (but no less relevant) is: Why should you, the part ic ipants in the Principles of Distr ibuted