Re: Email Web of Trust - some actual experience

Yakov Shafranovich <[email protected]> wrote:
> But how would this method is better than other methods for addressing 
> the hijacked machines problem? What are the differences?

  The web of trust doesn't address hijacked machines that well.
Rather, it explictly addresses long-lived machines sending email to
each other, and just happens to exclude DUL machines & trojaned ones.

  If an MTA is trojaned, then the web of trust can update it's trust
of that MTA.

> Wouldn't that put any new MTA in a catch-22 situtation: unable to send 
> mail until its sufficiently known and unable to be sufficiently known 
> since it cannot send mail?

  Yes, which is why one part of the design should be to permit
connections from IP's outside of the web, but accept only low rates of
messages.

  Alternately, the web of trust can be updated by hand.  example.com
register example.net, and lists it in the example.com web of trust.

> Of course that depends on how the information from this "web of 
> reputation" is used by receivers. If this information is used to reject 
> email, than it would be a problem. On the other hand, if this 
> information is only used to whitelist or give higher ratings to MTAs in 
> filtering, it is not as problematic.

  My suggestion is that the web of trust be used as nothing more than
one more piece of information into a local sites decision about the
message.  If the web of trust says "mostly trusted", the message can
go through lax content filtes.  If the web of trust says "mostly
untrusted", it can go through stringent content filters.

> Of course, past experiences with blacklists and how they are used,
> don't sound too promising.

  Blacklists are binary "yes/no" lists.  That was OK 4 years ago, but
isn't today.

  Blacklists are lists which expect users of the list to reject
messages from IP's in the list.  This also is not sufficient today.

> Assuming that they will be used for whitelisting. I can see such web 
> being used for blacklisting as well - if you are not in it, then we will 
> not accept email from you.

  That's a local decision, and is always possible.

  However, the design of the system should be such that people are
encouraged to make better decisions than simple rejection.

  Given the political history of blacklists, any new system which
isn't designed to be immune to political attacks will be sure to fail.

  Alan DeKok.