On 17/11/2008 15:35, "John Levine" <asrg(_at_)johnlevine(_dot_)com> wrote:
I still think there's some interesting research to be done about what
you'd do with reliable sender identities. It's easy enough to
whitelist mail from people you already know, but is there anything
clever one might do with an identity that you can trust to be stable
but you know nothing else about?
Probably not much. But if you know only a /little/ about them -- and can be
certain that the "who" you know about is the "who" who sent the message (as
measured by a particular authenticatable identifier), there's a lot you can
do.
For example, I suspect everyone here is familiar with the forwarding
problem: forwarders or mailing lists may inadvertently forward spam, and ISP
reputation systems assume the forwarding IP is a spammer. But if the
forwarder signed the message, and the ISP could do a VBR (or similar) check
to confirm that the forwarder really is a member of the "forwarder" class
(as determined by some sufficiently trustworthy external process), then the
ISP may choose to apply a different reputation algorithm.
This is one of those revolutionary ideas that a lot of people have thought
up over the years, and failed to implement because the identifier wasn't
sufficiently trustworthy.
Now, to be sure, the key could be stolen or the forwarder could get hacked
or all sorts of other things could happen which are outside of the scope of
DKIM or anything similar. But that's not a problem for modern reputation
systems (or even SpamAssassin), where it's possible to give a positive
weight without giving a free pass forever. If the behavior changes, the
reputation changes -- and the alternate algorithm or positive weight will be
overridden.
_______________________________________________
Asrg mailing list
Asrg(_at_)irtf(_dot_)org
https://www.irtf.org/mailman/listinfo/asrg