Trent Adams writes:
> To that end, some folks seem to be indicating that we should do that
> additional work here, within the DMARC WG.
Yes, at least I think there is work we can do to make ARC better, and
I still find it useful. The main problem in the ARC is the C part of
it. It was never and will never be a chain. Solution that requires
every single node who processes email to be updated simply will not
work in real world. This is the reason I do not belive DKIM2 will help
at all.
It is also the reason why NAT boxes spread much fater than IPv6. NAT
boxes can be implemented in single node, while IPv6 required that
every single router to be updated.
> Here’s my suggestion… if this group agrees that the current ARC
> experiment (as defined by RFC 8617) doesn’t hit the exact target
> (but may still have some value), I’d suggest that everyone who wants
> to see the learnings from this experiment implemented… that they
> join the DKIM2 work. The collaborators are keenly aware of what
> worked (and didn’t) in ARC and are baking into the DKIM2 proposals
> the best of what worked (and trying to correct what didn’t).
The DKIM2 also tries to make a chain of things, and tries to force
everybody in the path to change. I do not belive this will happen,
thus I do not think DKIM2 will be a solution that provides anything
useful in near future.
> With that, here’s my request to this group… I think it’d help for folks to
> consider the following when posting to the list:
>
> • Do you think that ARC has no value at all and should simply be buried?
No, I think ARC has value, even as specified now.
> • Do you believe there's “something” of value in the ARC model (even if it
> falls short)?
There is more useful in the ARC format and intermediate processin than
the actual Chain model of things.
> • Do you have concrete proposals for improving upon RFC 8617?
Remove the chain aspect of the ARC, i.e., change it to say that this
is single devices view of authentication of the email when it
processed it. As was pointed out in some other emails there is some
parts of authentication methods missing, but actually some programs
already do that, i.e., they add following ARC headers:
ARC-Authentication-Results: i=1;
ORIGINATING;
auth=pass [email protected] [email protected]
I.e., this tells that how this message was authenticated at that
point. Other similar missing authentication could be trusted-ip, etc.
This also means that we do not try to create a chain of ARC headers,
the per user filtering can use any information in the ARC headers
while processing the email, and if the signature of them validates, it
can trust the authentication results from that node depending what the
user have configured it to trust.
IKI ry members (30000 members) can trust that iki.fi has properly
checked the password of the authenticated smtp and that the from
address used matches that authenticated user, as they do have trust
relationship with iki.fi.
Globally you should not do such setting as you do not know what iki.fi
is. On the other hand if you have user in your system who gets lots of
emails from iki.fi users, they can probably configure their filters to
recognize that ARC-Authentication-Results from iki.fi.
So ARC do have value as building block for per user filtering, and it
also provides excellent debugging aid when you are trying to debug
issues. We quite often check ARC heaedrs of emails going through
iki.fi when someone complains that things did not work, and if we see
that ARC headers added by iki says SPF=pass, but DKIM=fail, or DKIM
did not exists at all, we do know that the most likely reason why the
next system did to accept the forwarded email is the missing DKIM
headers.
For lots of uses the ARC headers are in better than the Received
headers, which did use for debugging before ARC headers existed...
--
[email protected]
_______________________________________________
dmarc mailing list -- [email protected]
To unsubscribe send an email to [email protected]