mail-vet-discuss
[Top] [All Lists]

[mail-vet-discuss] draft-kucherawy-sender-auth-header-18.txt

2008-12-03 20:15:04

In today's draft, section 2.3  manages to increase the overstatements  
regarding path registration.  A path registration process is NOT an  
authentication process!  The term Authorization should be included  
whenever describing path registration methods.  An impasse remains.    
Once a domain has authorized an MTA, there will be those who want to  
consider an email-address as having been authenticated, whether  
located by a proprietary PRA algorithm or by the MailFrom, whenever it  
matches against an authorizing domain.  Why does this draft repeatedly  
describe authorization as being authentication?  Does repeating this  
lie over and over again in the draft somehow make it true?

This flawed logic puts recipients at significant risk.  Negative  
reputations necessary to defend against abuse MUST be applied against  
authenticated identifiers.  The PRA or the MailFrom are NOT  
authenticated, and thus will not safely block abuse.  Although some  
providers may disagree, the IP address of the SMTP client is the  
identifier that can be safely assigned negative reputations in  
conjunction with path registration exploitation.  The issue is not  
about the method used to determine reputations, only that the IP  
address of the SMTP client is absolutely essential for vetting the  
results obtained from path registration.

The deliberate exclusion of the SMTP client IP address was likely done  
to ensure this header, in the case of path registration, misleads  
recipients into trusting messages more than would be safe.  Lacking a  
means for MUA plugins to locate the SMTP client IP addresses, there is  
no recourse to properly deal with path registration exploitations.   
There is no reason why the _only_ authenticated identifier intimately  
associated with path registration, the _authorized_ IP address of the  
SMTP client, should be excluded. The header is called Authentication- 
Results and not Authorization-Results.

Stating Sender-ID or SPF as the method applied does not sufficiently  
define the scope of the path registration.  There was never an  
agreement reached as to the scope of the SPF record.  The  
authentication-results header fails to return an implied or expressed  
scope statement that may have been deduced from the path registration  
records, or that remains in an unknown state.  Without an implied or  
expressed scope, a process can not know what an authorizing domain  
intended to ensure, the MailFrom or the PRA.  When the scope is  
expressed as being for the MailFrom, it would also be dangerous to  
conclude the authorizing domain also ensures they control what the  
authorized address space emits.  Their motivation may have been to  
squelch bogus NDNs from other sources as a thrifty solution, without  
also going through the expense of imposing specific MailFrom  
restrictions on the MTAs being used.  In other words, only a negative  
result of path registration would be of limited benefit. :^(

An average person examining the authentication-results header is  
unlikely to have read that this draft has stipulated that only filter  
and annotation programs should view this information.  Nothing said in  
the draft will remedy the misleading header in the case of SPF and  
Sender-ID, especially while it also fails to provide any assured means  
to determine the SMTP client IP address.  One should ask, why does  
nearly every MUA offer a simple means to view email headers if users  
are not expected to view them?  Apple allows users to select any  
arbitrary header for display.  Unlike other trace headers, where there  
might be dozens, there should be only a limited number of  
Authentication-Results headers per message.  As such, it might become  
common to have this header displayed if it proves helpful in  
identifying messages likely to be bogus.  No plugins would be needed.

While path registration may indicate a message is from a suspect  
source, it remains dangerous to speculate what a positive result might  
imply.  It does NOT imply authentication.  There is nothing in this  
header to inform recipients, filters, or MUA annotation plugins what  
speculation might be safe, and whether it might be in regard to the  
MailFrom or the PRA. The misleading header label and missing SMTP  
client IP address information, in the case of SPF or Sender-ID,  
represents a dangerous and misleading overstatement as to the value of  
this header's results.  Do not regard authorization as being  
equivalent to authentication.

If this header is not to be viewed by average users, where the label  
and contained terms are currently misleading to anyone who understands  
the meaning of authentication, this can be repaired in simple  
fashion.  Dave Crocker's concern as to what is meant by "fail", rather  
than "neutral", can also be resolved in a similar way.  Use result  
codes rather than English terms.  It would make this header more  
internationally understood, easier for a program to interpret, and far  
less misleading to average users.  Safety first.

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

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