Re: DMARC: perspectives from a listadmin of large open-source lists

["Robin H. Johnson" <[e-mail]>]

On Tue, Apr 08, 2014 at 12:21:46AM -0400, John R Levine wrote:
> You would have to track which forwarders are well behaved and add
> valid X-O-A-R headers, but if you can do that, you can skip the header
> analysis and just whitelist the mail from the well behaved forwarders.

The XOAR proposal does specify:
| The X-Original-Authentication-Results header is only useful if the
| forwarder is trusted.  The forwarder is free to modify the headers and
| body of the message however it wishes and can generate new signatures
| over arbitrary X-Original-Authentication-Results headers.  Thus, the
| user SHOULD only trust X-Original-Authentication-Results if the message
| was delivered by known good forwarders, and forwarders SHOULD NOT
| propagate X-Original-Authentication-Results unless the previous
| forwarder is known to be good.
|
| For the purposes of this memo, a message was delivered through trusted
| forwarder if:
| - The DKIM signature passes
| - The DKIM domain is a trusted forwarder

I think the original scenario you described could be implemented by bad
players as follows:
- set up a mailman instance with DMARC support, that forges the XOAR
  header.
- Ensure that the mailman outgoing mail passes SPF+DKIM for the domain
  in question.


> Note that there are also well behaved things that don't pass DMARC and 
> don't have any original authentication results to report, with the usual 
> examples being mail-an-article at the NY Times and Wall Street Journal.
Those uses shouldn't be considered valid, and NYTimes has already moved
away from that, at least as of my test 5 minutes ago.
| MAIL FROM:<[e-mail]>
| RCPT TO:<[e-mail]>
| DATA
| ...
| From: robbat2 <[e-mail]>
| Sender: [e-mail]
| To: [e-mail]
| ...

> > The problem described WILL vanish when all mailing list apps implement
> > DMARC, but until then, it's really broken.
> Mailing list apps can't "implement DMARC" other than by getting rid of 
> every feature that makes lists more functional than simple forwarders. 
> Given that we haven't done so for any of the previous FUSSPs that didn't 
> contemplate mailing lists, because those features are useful to our users, 
> it seems unlikely we'll do so now.
By implement DMARC, I meant implement XOAR headers; VERP is too useful
to running lists to get rid of. Non-VERP bounce messages are still too
generic, even in this modern day.

-- 
Robin Hugh Johnson
Gentoo Linux: Developer, Infrastructure Lead
E-Mail     : [e-mail]
GnuPG FP   : 11ACBA4F 4778E3F6 E4EDF38E B27B944E 34884E85