To tell you the truth, I have always assumed that publishing code under a licence meant that that particular version of the code was under that licence and that it could never be revoked (unless the user violated it, of course). I didn't know there was a discussion around that fact.