Re: [dmarc-ietf] PSD simplification

[Dave Crocker <[e-mail]>]

On 12/12/2018 8:21 PM, Scott Kitterman wrote:
> If we extend DMARC with PSD and no limitations on PSO
> participation, then those considerations will apply to every domain that does
> not participate in DMARC (because the PSO can now get the data


Scott,

Thanks for the clarification.

Now I understand the nature of the privacy concern:

      (Public) registries that publish a DMARC record will get email
      traffic reports of their customers who do not publish DMARC
      records.

This does indeed seem a new attack surface, since the other powers of 
the registry mostly fall under denial of service rather than traffic 
analysis.

So the provisions of Section 6.1 in the -psd draft are meant to ensure 
that entries in the PSD registry are limited to any registry that:

>    adequately describes PSD policy to require domain owners to use DMARC
>    or that all domain owners are part of a single organization with the
>    PSO.

I suspect that this will have cases that prove rather more challenging 
to the Expert Reviewer than one might expect -- definition of 'single 
organization' can get sticky in some cases -- but still, the nature of 
the requirement is clear:  ensure that registry policy is documented and 
targets DMARC usage explicitly.

(Small point: Given this purpose for the registry, it's probably 
important for the registry's policy document to be publicly available, 
but that's not stated as required.)


Hmmm...

Let me suggest a much easier hack, which differs in utility mostly by 
being post hoc rather than the current draft's pre-hoc mechanism:

      Require the registry to publish another DNS record, in its _dmarc 
node, which a) asserts either than DMARC is required or that the subtree 
is part of a single organization, and b) contain a URL to the 
documentation for this.

A query for the DMARC record of the registry will also deliver this 
information record.  (This might be the first case in which the problem 
of getting 'other' TXT records is actually a feature and not a problem...)

That makes the information public, while avoiding the considerable 
overhead and problems of a new registry -- nevermind one that needs 
real-time querying.

d/


-- 
Dave Crocker
Brandenburg InternetWorking
bbiw.net