One. Last. Message. Of mine.

And sorry for all this mostly off-topic noise.

Steffen Nurpmeso wrote in
 <20240306214948.V5gSjSiU@steffen%sdaoden.eu>:
 |Steffen Nurpmeso via Postfix-users wrote in
 | <20231030191124.5ou-x%[email protected]>:
 ||It seems to me there is not much interest of mail operators in
 ||stepping to ed25519, reducing the payload of DNS and email?
 ||I know dkimpy supports it (and more -- but is python, uuuh!) for
 ||long, but OpenDKIM is unchanged for eight years.  (At least my
 ||sf.net import from 2017-09-23 still stands.)
 |
 |So now that i have DKIM myself i tested.
 |And *no* verification software i can reach actually supports
 |Ed25519-sha256 as of RFC 8463 from September 2018!
 |It is even *worse* than that.
 |
 |  - Google: at least reaches out to the RSA signature and verifies
 |    that, it ignores the other one saying "no key".
 |
 |  - Microsoft: fails the DKIM test if a RFC 8463 signature is
 |    present, no matter whether first or last!!!
 |    Is this *really* true?  That is really bad.

      + It even actively fails SHA1 DKIM signatures.
        I know these are deprecated, but if i use a rsa-sha1 and
        a rsa-sha256 signature in that order:

      Authentication-Results: spf=pass (sender IP is 217.144.132.164)
       smtp.mailfrom=sdaoden.eu; dkim=fail (body hash did not verify)
       header.d=sdaoden.eu;dmarc=bestguesspass action=none
       header.from=sdaoden.eu;compauth=pass reason=109

        The *very*same* message/-checkum passes Google:

      Authentication-Results: mx.google.com;
       dkim=pass (test mode) [email protected] header.s=lemon 
header.b=meYlPkTE;
       dkim=pass (test mode) [email protected] header.s=citron 
header.b=Cehr1W9z;
       spf=pass (google.com: domain of [email protected] designates 
217.144.132.164 as permitted sender) [email protected]

        Looking at that.  Say, the Microsoft
        Authentication-Results: does not denote its own domain
        name, no?  Ie i could not strip it.  I have not read RFC
        8601 for very too long to know, though.
        They do not look at the h=sha1 of the DNS record, do they.
        They do not look at the a= of the DKIM signature.

 |  - The software this list uses (rspamd i think): fails if the
 |    Ed25519 signature is first, aka does not reach out.  (Which it
 |    should, says DKIM, does it.  The DKIM standard is
 |    *fantastic*!)  It at least succeeds if the RSA is first.
 |
 |What a mess.  Even though explicitly envisioned in the DKIM
 |standard, it seems to me one cannot simply create two signatures,
 |as i wanted to do.  (For a while, at least; until i see Ed is
 |supported anywhere.  I had no plan, actually.)
 |
 |So as of today DKIM interoperability seems to mean:
 |
 |  - Place a single signature.
 |
 |  - It must be RSA-sha256.

And exactly only that.

 |RFC 6376 surely would have deserved something better.

Good night, greetings, and
Ciao from Germany,

--steffen
|
|Der Kragenbaer,                The moon bear,
|der holt sich munter           he cheerfully and one by one
|einen nach dem anderen runter  wa.ks himself off
|(By Robert Gernhardt)
_______________________________________________
Postfix-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to