Economics of SPAM [Was: Micorsoft's Sender ID Authentication......?]

Dave Crocker dhc2 at dcrocker.net
Sat Jun 11 18:56:06 UTC 2005


>   1a) must be simple so that many million server administrators can
>  understand it.
>   1b) must scale to millions of legitimate mail servers.
>   1c) must not break common functionality for users.

Good list.

To repeat the cliche, spam is a social problem.  Technical solutions can only 
follow social decisions.  Otherwise, we get technology dictating social policy. 
As bad as that is as a general rule, it is particularly bad for anything involve 
large-scale human communications, since the unintended consequences are certain 
to be massive and massively bad.

Spam (and virus attacks) seem particularly strong requirements for a layered 
defense, some proactive and some reactive.  Some involving authors and some 
involving operators.

Being able to white- or black-list an operator legitimately is particularly 
powerful.  They represent an aggregation of users and traffic.  So the leverage 
is enormous.  Perhaps because the payoff is so high, the dangers of 
mis-assignment are also huge.  So such listing needs to be done conservatively, 
which leaves lots of traffic unassigned.

Being able to white-list authors is equally spiffy.  In general, formulating a 
positive trusted core of communicants well might permit high quality service for 
relatively low costs, such as little or no content analysis, with its attendance 
statistical failings (false positives).

And so on...


  d/
  ---
  Dave Crocker
  Brandenburg InternetWorking
  +1.408.246.8253
  dcrocker  a t ...
  WE'VE MOVED to:  www.bbiw.net





More information about the NANOG mailing list