[as participant]

As I've mentioned before, the idea of doing forensic reporting of messages that 
fail DKIM verification has been a very valuable tool in terms of 
interoperability testing between DKIM implementations.  It's got value as well 
in showing signers how their mail is being altered in transit even when the 
signer and verifier are working properly (i.e., unanticipated mangling en 
route).

What's in draft-ietf-marf-dkim-reporting is essentially what I've developed and 
implemented almost since the beginning of the DKIM effort.  The DKIM working 
group didn't pick it up, but since the forensic reports themselves are actually 
ARF-based, I brought the work here.  It used to be part of 
draft-ietf-marf-authfailure-report but was broken out into its own document at 
(I think) our meeting in Maastricht.

I haven't had much in the way of feedback on it except through targeted 
requests for reviews.  When I finally got those, the suggestions I got were 
radically different from what I've implemented.  Given that, I'd like to invite 
the working group to consider it from the ground up in order to build something 
that has consensus.

So let's talk requirements.  The idea is to enable a signer to request that a 
verifier, on finding a message that fails verification testing, send a message 
back showing the canonicalized forms of the header and body, the idea being 
that the signer would keep a copy of the same and thus be able to see what 
changed in transit.  The mechanism of advertising that this is wanted needs to 
be simple and not add much burden to the existing system.  (That's why I 
grafted it on to the key record in the DNS, which has to be retrieved anyway.)  
The authfailure-report draft creates the syntax for the report, so this draft 
only needs to be able to communicate the request for the report and its 
parameters.

There are some obvious security considerations, such as being able to cause a 
storm of report mail if a lot of mail goes out all of which fails to validate.  
The interval stuff is in there to make it possible for the signer to mitigate 
such damage.

There's an adjunct ADSP angle, which requests a report if ADSP says "this 
should be signed by the author domain" and it wasn't, even if all the 
signatures passed.  The same requirements apply.

I think it's really as simple as that.

So if you don't like the dkim-reporting draft as it's written, how would you 
design something within those requirements?

I think Steve and John commented on this before.  I'll go dig up their old 
emails and comment on them as well, but I'll try to do so on this thread so 
it's all in one place.

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

Reply via email to