Re: [openpgp] Expected client behaviour ambiguity in signature verification

[Daniel Huigens <[e-mail]>]

go-crypto currently doesn't move signatures from one object to the other
if they're incorrectly ordered. I think it would return an error in
that case, since it verifies signatures during parsing (except for
third-party signatures on the User ID) - whether that's desirable is
of course also debatable but that's what it does currently.

Best,
Daniel


------- Original Message -------
On Tuesday, July 12th, 2022 at 19:38, Andrew Gallagher <[e-mail]> wrote:

> On 8 Jul 2022, at 15:22, Daniel Huigens [e-mail] wrote:
>
> > On Friday, July 8th, 2022 at 15:14, Justus Winter wrote:
> >
> > > You said before that you prefer not to use the reordering heuristic.
> > > But how else should the keyserver fix the certificate?
> >
> > I meant more that they could do a manual intervention for the keys where
> > that happened, perhaps. (Particularly in the case where the keyserver
> > itself broke the key, I didn't mean that they should fix up these keys
> > in general.) Then, they could also verify the signature packets with
> > each of the components, to check where it should be placed.
>
>
> Hockeypuck already fixes such badly-ordered sigs, because it uses go-crypto to parse the key material before serving to clients, and AIUI this normalises signatures implicitly. (There are knock-on issues about how to deal with the non-normalised back end storage but those are orthogonal to this discussion…)
>
> A