[openpgp] Re: Persistent Symmetric Keys: new algorithms or new packets?

Justus Winter <[email protected]> Wed, 23 Jul 2025 16:18:28 +0200

Newsgroups	gmane.ietf.openpgp
Message-ID	<[email protected]>

Daniel Huigens <[email protected]> writes:

> On Wednesday, July 23rd, 2025 at 15:15, Justus Winter wrote:
>> I think wouldn't mind the new kind of key packet also using PKESK
>> packets and signature packets.
>
> How do you envision this part concretely? Those packets have a public
> key algorithm field, so we would presumably still need to register at
> least one algorithm ID for this scenario, then?
>
> Or we could set it to 0, which is reserved?

I would still register public key algorithm IDs for AEAD and HMAC keys.

I'm less worried about having a few symmetric algorithms in the "public
key algorithm" registry than about re-using the key packets, which have
a public key and a secret key variant, and the former doesn't make sense
with symmetric keys, yet the latter is defined in terms of the former,
and because of that relationship current implementations may assume that
one can derive a public key packet from a secret key packet.

And then there is the expectation of being able to compute fingerprints,
and have them be reasonably unique.

All of those concerns just go away when we use a different key packet.
And, I think implementations then have a chance of adding a new
interface for these packets (on the low-level, I'm not suggesting
exposing this as a new interface to users, though I'm sure that these
new mechanisms will require application-specific handling).

Best,
Justus

_______________________________________________
openpgp mailing list -- [email protected]
To unsubscribe send an email to [email protected]

signature.asc (application/pgp-signature, 584 B)