ietf-dkim
[Top] [All Lists]

Re: [ietf-dkim] New Issue: protecting a domain name vs. protecting a domain tree

2008-04-10 13:01:42

On Apr 9, 2008, at 11:43 AM, MH Michael Hammer (5304) wrote:
In response to the question Dave asks, I like the idea of providing  
the option of protecting an entire (sub)tree within in a domain. My  
question to the gurus is whether there is a clean way to identify  
"main" domains below a TLD. For the generics this would appear to be  
straight forward. For country TLDs I'm not so sure. Some country  
TLDs might always require a .co.TLD or .edu.tld (or something  
similar). Not only is there inconsistency across such TLDs, there  
may be inconsistency over time as far as requirements within a TLD.

What started me thinking along this line was allowing a base domain  
(if you will) to make an assertion that ALL subdomains only send  
signed mail (or never sign mail or ?)

Technically, Dave Crocker is right.  The issue he raises is valid  
since message content is independent of SMTP.  When taken to heart,  
one can not expect _any_ DNS records relate to what might be email- 
addresses contained within the messages.

One simple assertion can overcome this issue.  Declare protections  
afforded by ADSP _only_ relate to email-addresses exchanged using  
SMTP.   With this statement, there would be MX or A DNS resource  
records required by the domain.  Presence of these records therefore  
offers a means to validate the domains.

To put an upper limit on the number of policy related DNS transactions  
that could increase over time, require publishing MX records with any  
SMTP policy record.  This would place an upper limit on the number of  
transactions needed to determine presence of SMTP policy.   The first  
step in evaluating SMTP policy would be requesting an MX record.  To  
determine whether the domain might be valid for SMTP, a subsequent  
transaction could check for A records.  Until A record discovery  
becomes deprecated, and to avoid tree walking, domains seeking  
protection should be required to publish ADSP records at every node  
there is also A records.  The existence of policy records in the  
absence of MX records would also refute any message comes from the  
domain.  As policy records would need to be published in conjunction  
with A records, the requirement that policy be qualified by MX records  
eliminates any need to also publish bogus MX records as have been  
suggested as an alternative strategy.

Acceptance of messages independent of SMTP delivery is separate matter  
not covered by ADSP.  This could be handled through a scope statement  
in the ADSP record, but this field is currently missing.  The way  
forward would be to declare that ADSP pertains to SMTP exchanged  
messages.  When handling a mixture of messages exchanged using SMTP  
and other protocols, there will be potential conflicts.  When the  
exchange protocol is not apparent to recipients, these messages may be  
seen as not complaint with ADSP assertions when a default assumption  
of SMTP is used.

-Doug 
   
_______________________________________________
NOTE WELL: This list operates according to 
http://mipassoc.org/dkim/ietf-list-rules.html

<Prev in Thread] Current Thread [Next in Thread>