Re: [openpgp] Put Signature in an Email's Header

[Bart Butler <[e-mail]>]

Yes, the idea was to have one multipart/mixed section and have potentially N signature headers inserted between marker position 1 and 2. There would be no modification or MTA insertion of anything else below marker position 2.

But after looking at your examples and thinking about it I think the encapsulation approach you suggested would be better, as then there’s a clear line between what is signed and what isn’t and I think that would be much cleaner and easier on implementors as we could maintain the existing logic from PGP/MIME for computing signature data over whole MIME sections. I think an additional MIME layer is a strength here.

On Wed, Aug 9, 2023 at 10:56 AM, Kai Engert <[e-mail]> wrote:

On 08.08.23 19:06, Bart Butler wrote:
> Why not take a page out of DKIM’s book and just put the signature
> header(s) in the protected headers directly, maybe at the top,
> except exclude only that header from the signed data if present? Seems
> like that would also be backwards compatible with existing PGP/MIME if
> you wanted to sign both ways as well.

You mean like the following?

The app creating the message would take this as the input for signature
calculation:

Content-Type: multipart/mixed; boundary="prot-hdr";
protected-headers="v1"
<all protected headers>
--prot-hdr
payload
--prot-hdr--

Then the signature is inserted ...

<outer headers>
Content-Type: multipart/mixed; boundary="prot-hdr";
protected-headers="v1"
-- marker-position 1
Header-Signature: micalg=pgp-sha256;
protocol="application/pgp-signature"; sigdata=
wsF5BAABCAAjFiEEIdFuZ...
=9HqM
-- marker-position 2
<all protected headers>
-- marker-position 3
--prot-hdr
payload
--prot-hdr--

If "multipart/mixed protected-headers" is the outermost content-type,
are MTAs allowed to add additional headers at
marker positions 1, 2 or 3 ?

Then it would ambiguous for the receiver to decide which data to use as
input for the signature verification.

Kai