[ietf-dkim] list expanders (was Re: chained signatures, was l= summary)

[Douglas Otis]

On Jun 22, 2009, at 2:51 AM, Charles Lindsey wrote:

> On Fri, 19 Jun 2009 17:55:57 +0100, Douglas Otis <dotis at mail- 
> abuse.org>
> wrote:
>
>> It dangerous to consider A-R headers of unknown origins as somehow  
>> inherently safe......
>
> Unless they are included in a signature.

Unverifiable DKIM signatures will not confirm the origin of an A-R  
header.

> An A-R record always includes an idication of the domain that  
> purported to have place it there. If it is signed by that same  
> domain (as would be the case in the scenarios we are discussing),  
> then more reliance can be placed on it (depending on your opinion of  
> that signer - but you opinion of the manager of a mailing list you  
> have subscribed to is likely to be quite high).

There is no assurance that RFC 5451 "authserv-id" have any  
relationship with an RFC 4871 d= values.  There is also no assurance  
that  bogus A-R headers would have been removed.  Amongst  the  
confused advice about retaining A-R headers, there is also not a  
certain practice of not signing A-R headers, even when A-R headers are  
not added.  Do you expect there will be d= requirements imposed upon  
authserv-ids, since that is not how A-R validation works?

> I agree that an unsigned A-R is dubious, but even then if it  
> purports to have been placed there by a domain which
>    a) the message has been passed through, and
>    b) you are prepared to trust to have removed any pre-existing  
> bogus A-R purporting to have been placed there by that domain
> then it should be pretty safe (and this was indeed the case for the  
> example we were discussing).


What this would be trusting may not be common practice.  In addition,  
there is no defined relationship between authserv-id and d= values or,  
for that matter, domains in general.

A mailing list might use a signing domain of "foo.tld" and the  
authserv-id of "bar.unique".

By allowing inclusion of A-R headers having unknown origins remains a  
risky practice since a recipient's MUA might trust these headers.

A safe practice would remove all foreign A-R headers.  Exceptions made  
when signed by verified DKIM signatures still requires trusting the  
DKIM signature not to introduce misleading A-R headers.  Again, there  
is no direct relationship between RFC 5451 "authserv-id" and RFC 4871  
d= values.

-Doug