[openpgp] Re: New Version Notification for draft-gallagher-e mail-unobtrusive-signatures-00.txt

Andrew Gallagher <[email protected]> Wed, 14 May 2025 14:22:23 +0100

Newsgroups	gmane.ietf.openpgp
Message-ID	<[email protected]>

On 14 May 2025, at 11:09, Stephen Farrell <[email protected]> wrote:
> 
> DKIM2 is also proposing an algorithm to "unwind" some of the
> changes that may occur to mail in transit (e.g. mailing list
> footers) so one can validate a signature from before those
> changes were made. If we create new mails with multipart/mixed
> those might interact with that still-being-developed algorithm.

So long as the message that is end-to-end-signed is the same one that gets signed by the first DKIM signer (i.e. the sender’s mail provider), then any subsequent changes and/or reversions in transit should break (or fix) both the DKIM and the E2E signature equally - which would also be the case for traditional PGP/MIME signatures.

The only caveat is whether DKIM’s relaxed/relaxed canonicalisation allows a DKIM signature over a slightly mangled message (e.g. trailing whitespace) to validate when an E2E signature would not - but the message canonicalisation guidance in section 5.4 of the current draft is designed to avoid this.

>> I personally feel some pain from people not understanding 'signature.asc'
>> attachments, but not a huge amount.
> 
> Ditto. Though it does "break" things when people say they
> can't read your signed emails because of that, so it is a
> real issue.

Personally, I don’t get many such complaints either. I hadn’t had one in a long time, but a few weeks ago I had a contractor complain because he was expecting me to send him some documents and was trying to unzip the signature attachment. I sent him the real documents and left it at that, but that was because I accepted the consequences when I enabled email signatures; someone who did not make such an informed choice might not be so accepting. We should always keep in mind the difference in practical outcomes between users making config changes that may cause breakage on an individual scale, and software making a change to the defaults that may cause breakage on a mass scale. Pain points that are tolerable in the first scenario would not be tolerable in the second.

A

_______________________________________________
openpgp mailing list -- [email protected]
To unsubscribe send an email to [email protected]

signature.asc (application/pgp-signature, 833 B)