Allowing "duplicate" signatures

David Shaw dshaw@jabberwocky.com
Mon Aug 6 16:08:01 2001


On Mon, Aug 06, 2001 at 09:46:47AM +0100, Matthew Byng-Maddick wrote:

> On Sun, Aug 05, 2001 at 07:02:50PM -0400, Michael Young wrote:
> > As it stands, GnuPG refuses to sign a key/name pair with a particular
> > key if a signature by that signing-key already exists. There are
> > several reasons that you might want to do that, though. As noted
> > in the code, the existing signature could be revoked. Similarly,
> > it could be expired. You might also want to generate a new
> > signature with new properties (subpacket values):
> > new expiration time;
> > new signature type (not yet selectable, but I'd like it to be);
> > different notation data;
>
> Yes, definitely. If the patches posted here that I've written or something
> similar gets included in GnuPG, then there is value in being able to add
> and revoke signatures with various different notation data in them.
A problem here is that the OpenPGP spec has no notion of revocation of a specific signature. A revocation is actually another signature that in effect says "I *DON'T* certify this key". There is no way to specify which of multiple possible signatures from a given key it is intended to revoke. Trust-wise, the calculation has to match up the valid sigs with valid revocations (the time stamp can help here to a certain degree) and see if there are more sigs than revocations. I suppose it could be said that if a given revocation signature matches the original signature in all subpackets except for the signature type being set to revoke (0x30) then the revocation should apply to the particular signature that was matched. That's out of my head, and not in the RFC though. David -- David Shaw | dshaw@jabberwocky.com | WWW http://www.jabberwocky.com/ +---------------------------------------------------------------------------+ "There are two major products that come out of Berkeley: LSD and UNIX. We don't believe this to be a coincidence." - Jeremy S. Anderson