On Mon, 2004-10-04 at 08:48, Dave Crocker wrote:
On Sat, 2 Oct 2004 21:05:49 -0400, John Leslie wrote:
Having a class/state assertion rather than a quality
assertion makes for a simpler solution. Allowing a
flag that indicates an affiliation to a group that
assures BCP regarding the sending of mail seems like
a good way to go, hence the term accreditation.
Doug makes a reasonable case for this extension. Dave, I
think, is less enthusiastic. How do others feel?
just to add this to the thread, for reference:
Accreditation is a new topic. It is so completely tied to human
and organization issues that I expect it to prove to be extremely
difficult to do a thorough job of. Also, I think there is an
important technical constraint, namely limitations on what is
reasonable to expect of a receiving SMTP server to perform in
real-time, when there is a potentially large flow of incoming
messages.
The affiliation indication offers a means to request consideration
before a filter is applied. I would not expect this work group to
define policies implicit in the indication, only that the indication be
accommodated as a first step.
I tend to use ATM/credit card information services as an exemplar
for mail accreditation. For real-time transactions, there is an
'approval' mechanism. It returns a yes/no. For major
transactions, there is a 'reporting' mechanism that supplies a
large amount of source material; the assessment of approval is
left to the requestor.
For real-time decisions by receiving SMTP servers, I believe we
should provide a very simple rating system. It should give an
explicit recommendation, rather than supply any sort of
descriptive attribute/value information about the domain name
under inspection.
Also, by hiding all of the information the rating service uses to
return a yes/no, and by keeping the mechanism so simple, there is
a chance we can ship something reasonably soon.
The current DNA proposal has 5 values, where 0 is 'don't know'.
That is more than yes/no, but not much.
A rating system is of no value for determining disposition of mail. It
provides no clear indication when to refuse mail nor what an 'A' rating
versus a 'B' rating means from one rating system to the next.
The "state" proposal rather than "rating" was to avoid the uncertainty
of appropriate action. The binary yes/no as used today avoids this
problem. Information regarding abuse changes by the hour. A good
provider may permit operation of an abusive account at varying times.
For reputation services that dynamically communicate with both the
service provider and the mail recipient, a responsive mode of operation
is possible. From a good provider, the vast majority of the mail is
acceptable, but a bad account can send a significant amount of abusive
mail until it is detected and disabled. The time to disable these
accounts can be accomplished within a few hours. Temporary refusals "as
needed" significantly reduces the abusive traffic allowed to pass during
this period. Providing an aggressive system with the _least_ amount of
uncertainty for implied actions would be as follows:
'A' Good Standing:..................... Accept
'B' Good Standing with current abuse:.. Refuse Temporarily
'C' New:............................... Accept with limits
'D' New with current abuse:............ Refuse Temporarily
'E' Bad:............................... Reject
Interpretation of these states is at the discretion of the recipient.
Refuse Temporarily could be Accept and the message receives more
filtering or offered a different filtering score. A turn-around on this
information will significantly impact the amount of abusive mail allowed
to pass through the system. This type of indication offers a warning
and logical response that elicits closer communications between
providers and reputation services and will significantly reduce the
abuse allowed.
The current draft is:
'A' Good Standing:..................... Accept
'B' Less than Good Standing:........... Accept/Reject?
'C' New:............................... Accept/Reject?
'D' Somewhat Abusive................... Accept/Reject?
'E' Bad:............................... Reject
It is not clear with such a rating system what differentiates each
rating level. Would an 'A' rating be indicative of retaining a 97%
level of good/abusive? Perhaps 'B' a 94%, and 'D' a 91%? Is this too
broad? Whatever level is selected, this level will be maintained by the
dedicated abuser.
-Doug