I was in the middle of typing up the interesting history of this paradox, which goes to the heart of frequentist v. bayesian probability when I realized that there was probably a wikipedia article. And there is:
Two envelopes problem - Wikipedia, the free encyclopedia
