Re: [openpgp] Weird OIDs in the 4880bis draft

[Daniel Huigens <[e-mail]>]

Hi all,

I support this change.

I don't want to muddy the waters too much, but I've also created [!242]
as an alternative to this, which defines new algorithm IDs for Ed25519,
Ed448, X25519, and X448.

Unlike Justus's MR, it rips out the OIDs for these algorithms, and the
scaffolding around them to put them in MPIs. So, it's a fairly large
change, but a lot of it is removing stuff.

For backwards compatibility with Curve25519 keys out there, people will
of course still have to support the Curve25519 and Ed25519 OIDs, but we
can refer to the previous crypto refresh for those, IMO.

As a more minor note, in my view this MR also replaces [!223] -
"Remove session key checksum and padding for v5 [soon v6] ECDH keys" -
in the sense that it doesn't add the checksum and padding in these new
algorithms, reducing the need to remove them from v5/v6 ECDH PKESKs.
That way, we also keep the invariant that the behavior of a public-key
algorithm doesn't depend on the version of the PKESK.

In other words, it basically would mean that (if we merge this, and not
!223) the crypto refresh would make almost no changes at all to ECDH and
ECDSA compared to RFC6637, except for the addition of Brainpool.
(In my personal opinion, we could even rip out ECDH and ECDSA entirely,
and keep referring to RFC6637 and the previous crypto-refresh for those
as well, but perhaps that's a bit too radical, so I'm not proposing it
as part of this :))

In any case, even without that, I think this would simplify the spec a
lot. That being said, if you all think this is too large a change, I'm
happy to withdraw it in favor of !242.

Best,
Daniel

[!242]: https://gitlab.com/openpgp-wg/rfc4880bis/-/merge_requests/242
[!223]: https://gitlab.com/openpgp-wg/rfc4880bis/-/merge_requests/223