Re: a Message-ID proposal
There were some obvious bugs in my proposal:
- The domain should be included in the Message-ID and in the string
that is hashed to make the signature.
- The way I proposed generating the uniquifier is utterly wrong. I
have some other ideas for generating the uniquifier but for simplicity
we could assume the uniquifer is either like the current scheme
(letter + pid) or random.
So the scheme looks like this:
Message-ID = timestamp . uniquifier . signature @ domain
uniquifier = letter pid | random
signature = checksum(secret . timestamp . uniqifier @ domain)