From: Kee Hinckley <nazgul(_at_)somewhere(_dot_)com>
...
Let me propose a rule for proposals.
No proposal without an explanation of the incentives for the senders
and receivers to adopt the system at three stages: 10%, 50% and 90%
deployed. And at each stage, explain what actions the spammers are
being forced to take.
That omits the most important stages at 0.01%, 0.1%, and 1% measured
in target mailboxes protected or total mail scanned. 0.01% is 50,000
mailboxes, which the size of a modest regional ISP or a large company.
It's not easy to convince the people responsible at such outfits to
risk wasting their time and effort on something that turns out to be
like almost all anti-spam solutions, either not worth the effort (<80%
effective) or fine if you don't care about false positives or scaling.
It's not easy even to get the first 0.0001% or 500 mailboxes protected
by months of flogging a solution in slashdot, freshmeat, spam-l, and
news.admin.net-abuse.email, but it is possible. It turns out that
people in those forums are more interested in talking about spam
solutions than installing them. It is quite difficult to reach 0.1%
or 500,000 mailboxes, because you must convince more than the handful
of anti-spam and anti-spammer hobbyists who can choose their own mail
software.
I would argue that a system that offers no benefits until you've
passed the 50% point is never going to be adopted.
That's optimistic by more than a factor of 10. Even 1% is probably
too high a threshold. Something that is useless until 5% has no hope
unless you do the impossible and convince an AOL to use a solution
invented outside. Even an AOL is only about 5%.
On the other hand. A system which is virtually guaranteed to fail at
the 90% point (challenge/response, possibly content filters) can
demonstrably shown to have early adopters because, while it may fail
when scaled, it works now.
That's 1000X optimistic. As far as I can tell, one body checksum
scheme has failed or been retired. There may have been other reasons,
but it was having scaling problems at 0.09% despite getting a lot of
attention from early adopter hobbyists.
Let's try and leverage those social forces. (My anthro advisor would
be proud :-)
Or at least admit they exist and be realistic about them.
Note that these claims are based not only on my prejudices but also
my experiencd with the DCC for the last two years. I figure the DCC
has reached 0.5% or conceivably a little more.
How many contributors to this list have any experience pushing spam
solutions at or above the 0.01% deployed level? What about 0.1%?
I don't want to offend, but to emphasize what I understand as part
of Kee Hinckley's point, that spam suffers from the talk-is-cheap
syndrome in more than one way.
Vernon Schryver vjs(_at_)rhyolite(_dot_)com
_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg