Popowycz, Alex writes:
 > Vlatko

Please don't feed the trolls (including me, I'm regretting my role
in this thread).  There's work to be done here, and Vlatko seems
uninterested in helping with it (eg, when asked for specific
references, he says "I'm not a search engine").

What I gather from Vlatko's posts is that there is a use case where an
entity (eg, a small business; called "ENTITY" below) wants its own
domain (called "OWNDOM" below) referenced in correspondence, but
prefers not to maintain a single presence (even as a VPS) on the
Internet.  Instead, ENTITY uses an ESP (below, "ESP" denotes
aparticular ESP) to send and receive mail, and ESP provides the usual
set of authentication services for its host.  The need is to provide
credentials and an authentication protocol for mailboxes in OWNDOM
that will satisfy identity alignment, and all actors will voluntarily
participate.

SPF would require that ESP permit ENTITY to specify MAIL FROM.[1]
Then ENTITY publishes an SPF record pointing to ESP's MXs and thus
provides identity alignment for OWNDOM.  However, this requires that
ESP trust that ENTITY has the right to use OWNDOM, and that may be
problematic.  Eg, spammers could sweep the DNS looking for SPF records
pointing to ESP and then specify a domain publishing such a record in
MAIL FROM.  Perhaps an additional protocol could be designed so that
registrars can vouch for entities (OAuth2?) but that's outside the
scope of this post.

DKIM seems to be easy, and requires no cooperation as long as ESP
passes mail through unchanged except for its own signature and trace
fields.  ENTITY just generates a keypair, publishes the public key
through its DNS, and DKIM signs its own messages with the private key.
This should be a SMOP (eg, Python has at least two packages on PyPI
for handling DKIM).  Identity alignment for OWNDOM will be satisfied.

An alternative would be for ENTITY to trust ESP a little bit, have ESP
generate the keypair, send the public key to ENTITY which publishes an
appropriate DKIM resource, and then the ESP DKIM signs for ENTITY (and
optionally for itself as well).  If this capability isn't available at
ESP, that too should be a SMOP (basically a database lookup to get the
key for ENTITY).  Identity alignment for OWNDOM will be satisfied.

This last approach seems like a business opportunity for ESPs.  Am I
missing something?

I don't understand why Vlatko thinks he has a problem that needs a
new protocol to solve.  What am I missing?


Footnotes: 
[1]  It's true that ENTITY could publish records in OWNDOM pointing to
the MXs as hosts (CNAME), but then HELO would be valid for any user of
ESP spoofing OWNDOM AFAICS.

_______________________________________________
dmarc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dmarc

Reply via email to