On Jan 23, 2012, at 2:18 PM, Murray S. Kucherawy wrote:

>> -----Original Message-----
>> From: [email protected] [mailto:[email protected]] On Behalf Of 
>> Shmuel Metz
>> Sent: Monday, January 23, 2012 10:50 AM
>> To: Message Abuse Report Format working group
>> Subject: [marf] draft-ietf-marf-as Section 5 Solicited and Unsolicited 
>> Reports
>> 
>> I believe that 5.  Solicited and Unsolicited Reports should list the
>> abuse address from the whois record of the source IP as a reasonable
>> candidate for receiving feedback.
> 
> I have some concerns about doing this, since the reply from a WHOIS query is 
> non-standard.  Do we really want to say "apply some unspecified heuristic to 
> the WHOIS reply to get that address"?

It's all heuristics. However, ARIN and RIPE at least have some structure in 
their responses, including explicitly recorded abuse contacts. When they exist 
(which they usually do) that does give you an IP address to abuse address 
mapping - and while the contact is often too far up the delegation tree to be 
the right contact, it's seldom an actively bad contact:

OrgAbuseHandle: ABUSE1036-ARIN
OrgAbuseName:   Abuse Department
OrgAbusePhone:  +1-510-580-4100
OrgAbuseEmail:  [email protected]
OrgAbuseRef:    http://whois.arin.net/rest/poc/ABUSE1036-ARIN

> 
>> If there is a PTR for that address, would an associated abuse address
>> be a reasonable candidate for receiving feedback? If so, would it only
>> be reasonable for FCrDNS?
> 
> I don't think so.  I don't think we want to start encouraging people to try 
> to find any domain to which to prepend "abuse@" to start sending reports.  
> DKIM is the exception, because a valid DKIM signature makes a strong 
> statement the likes of "Yes, we handled this message."  A PTR record, for 
> example, does not.

DKIM states that the owner of the d= hostname signed[1] the message. That could 
mean anything between them being the spammer, to them being the ISP of a end 
user with a compromised box. Whether that's an appropriate entity to contact 
requires applying some heuristics, and mapping that d= value to an appropriate 
email address requires some more.

In the case where an ESP is signing the mail sent by their customers with a d= 
value inside the customers domain then there may not be an abuse@X address for 
any given d=X or d=Y.X, and even when there is it's unlikely to be the right 
address to contact for abuse issues. Simple heuristics based on ARIN or RIPE 
registry data are likely to be much more accurate in that case.

Cheers,
  Steve

[1] Or delegated signing responsibility to, or...
_______________________________________________
marf mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/marf

Reply via email to