Everything else aside, the recent Wikileaks/Guardian fiasco (in which the passphrase for a widely-distributed encrypted file containing an un-redacted database of Wikileaks cables ended up published in a book by a Guardian editor) nicely demonstrates an important cryptologic principle: the security properties of keys used for authentication and those used for decryption are quite different.
Authentication keys, such as login passwords, become effectively useless once they are changed (unless they are re-used in other contexts). An attacker who learns an old authentication key would have to travel back in time to make any use of it. But old decryption keys, even after they have been changed, can remain as valuable as the secrets they once protected, forever. Old ciphertext can still be decrypted with the old keys, even if newer ciphertext can't.
And it appears that confusion between these two concepts is at the root of the leak here. Assuming the Guardian editor's narrative accurately describes his understanding of what was going on, he believed that the passphrase he had been given was a temporary password that would have already been rendered useless by the time his book would be published. But that's not what it was at all; it was a decryption key -- for a file whose ciphertext was widely available.
It might be tempting for us, as cryptographers and security engineers, to snicker at both Wikileaks and the Guardian for the sloppy practices that allowed this high-stakes mishap to have happened in the first place. But we should also observe that confusion between the semantics of authentication and of confidentiality happens because these are, in fact, subtle concepts that are as poorly understood as they are intertwined, even among those who might now be laughing the hardest. The crypto literature is full of examples of protocol failures that have exactly this confusion at their root.
And it should also remind us that, again, cryptographic usability matters. Sometimes quite a bit.