This paper is related to, but not equivalent to, the keynote talk given at the workshop on Physics of Computation.. This published version concentrates more on matters of detail, and is not intended as a survey.
