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

[Wiktor Kwapisiewicz <[e-mail]>]

On 14.02.2023 14:32, Peter Gutmann wrote:
> Wiktor Kwapisiewicz <wiktor=[e-mail]> writes:
>> MPIs being counted in bits not bytes,
> That actually leads to an interesting point, I realise that the bit-lengths
> are too ingrained to change but just out of interest, is there any
> implementation that actually uses the bit lengths, as opposed to putting a * 8
> in every write and a / 8 in every read?

I'm not aware of any but I guess the fastest way to get answer to this 
question would be creating a test case here:
https://tests.sequoia-pgp.org/

MPI count is a rabbit-hole, e.g. the RFC says the count starts from the 
first bit set (1) so in theory the MPI could be padded (prefixed) with 
any number of clear (0) bits (or bytes):
   https://www.rfc-editor.org/rfc/rfc4880#section-3.2

If I'm reading the draft well it seems to be fixed in:
   https://openpgp-wg.gitlab.io/rfc4880bis/#section-3.2.1

"followed by the smallest series of octets that can represent the value 
while stripping off any leading zero octets".

But it seems this discussion has strayed a little bit too far from it's 
original goal ;)

Kind regards,
Wiktor