Signing our keys

[hughes@ah.com (Eric Hughes)]

>If we are willing to sign a key based on an
>entity that we KNOW does not really exist, then what
>does a signature mean? 

The key is the identity, period.  Let's get that straight up front.
Signing a key does not change the identity.

Signatures on keys are attestations that the key belongs to some
person or email address.  Signing a key attests that an identity has a
name.

I've developed a criterion for signing pseudonym keys.  The only party
(other than the holder) who can gain any certain knowledge that an
e-mail address maps to the holder of a given key is the provider of
the e-mail address.  In other words, Julf's server should sign
wonderer's key.  

The following applies to any system providing pseudonyms, i.e. mail
addresses.  The mail server should have a public key.  The public key
of a user would be encrypted with the server's public key and
forwarded to the server.  The server would accept as authentication of
this public key the same authentication that it accepts for everything
else.  Once it has the key, it can sign it and return it to the
individual, who can then publish it.

>Something strange is going on in the word when nym's
>are signing each other's keys...

The one pseudonym is attesting that they reach the same individual
each time they send mail to the other pseudonym encrypted with the
public key claimed by that other pseudonym.  This attestation is not
as strong as person-to-person contact, but as long as the signer
reasonably believes that mail delivery system functions as it claims
to, i.e. no interposers, the signature does actually mean something.

Eric