Re: [openpgp] v5 interoperability

[Heiko Schäfer <[e-mail]>]

On 4/12/24 19:28, Paul Schaub wrote:
> I recently took a look at the LibrePGP document and found some more 
> things that would result in  inconsistent behavior.
>
> CR states that only v6 keys shall create v6 signatures. LibrePGP on 
> the other hand allows any key to create v6 signatures.
> It also allows v6 sigs to have zero length padding (the way this ought 
> to be encoded is imho underspecified), while CR binds padding length 
> strictly to the digest algorithm.
>
> I'm sure there are more of these inconsistencies which will make 
> interop and parser design a pain...

I share these concerns, but wonder if Kai's suggestion adds any pain to 
what we're already looking forward to, on the parser front, without 
Kai's suggestion:
An ecosystem that will possibly contain artifacts with formats that may 
not be reliably parsable based on just the version number of a packet.

For example: Let's imagine GnuPG were to emit "slightly different v6 
signatures" in the future. Parsers will then have to deal with that 
format (if only just by rejecting the signature).

At least on the parser side I struggle to imagine how mixing versions 
(as per Kai's mail) makes things worse (which is not to say that I think 
there are *no* serious problems with the idea).