On 11/04/2013 11:02 AM, MFPA wrote:
And as an aside, does it really make a difference to only sign some
UIDs and not others? Does GnuPG actually take account of which UIDs
are signed in its validity or trust calculations?
Yes, it does make a difference.
Let's say I make key X and attach to User IDs to it:
* Daniel Kahn Gillmor <d...@fifthhorseman.net>
* Alice Munroe <al...@example.com>
You meet me, check my identity, verify that i'm actually dkg, and just
sign the first User ID (because you have been unable to verify whether i
am also somehow Alice Munroe). (in fact, i am not Alice Munroe, but i
would like to be able to read her mail)
At some point, you find you want to encrypt a message to Alice Munroe
(who you met at a conference, perhaps). If you had certified both User
IDs on my key, gpg would be happy to encrypt the message to my key
instead of Alice's actual key. If i get a copy of that message, i would
be able to read it. This would be bad.
An OpenPGP certification (a "keysigning") is an identity assertion, over
*both* the key and the User ID. It says "this key K belongs to the
person known in the real world by the User ID U", and it is
cryptographically signed by the person making the assertion.
If you substitute some arbitrary other User ID for U, the meaning of the
certification changes radically (and the cryptographic certification
breaks). This is an intended feature.
--dkg
_______________________________________________
Gnupg-users mailing list
Gnupg-users@gnupg.org
http://lists.gnupg.org/mailman/listinfo/gnupg-users
