Right. The dialog with the crackers, showing how they use Markov chains, recombination of already-found passwords, and other adaptive techniques was new to me.

Right. The dialog with the crackers, showing how they use Markov chains, recombination of already-found passwords, and other adaptive techniques was new to me. They were able to turn up lots of seemingly obscure passwords very fast. Basically, the strength of passwords can look really good on paper (96^12 possibilities or whatever), but in practice the entropy is much less: individual characters are not as independent as we'd like to think, and your clever word or phrase may be in a dictionary already (or may be a short Levenstein distance from something in a dictionary).