equitable counting

Alexy Khrabrov braver@pobox.com
Mon Oct 20 22:42:42 UTC 2003

I'm a scientist, so I ask a question which may be theoretical,
but it's about the DCC design: should counting be equitable?  

We detect spam in the most rigorous way: spam consists of
messages which (allowing for personalizations caught by fuzzy
checksums) were seen in bulk.  In order to get that "many"
count from a DCC server, you have to allow "many" previous
clients to raise the count.  The first < threshhold, say N,
clients will not get the benefit of the counting.  The only
equitable solution seems to be greylisting with counting,
which is what seems to be implemented in the new DCC version.
Is is why greylisting is included with DCC?  What other
equitable solutions are possible?

This question is theoretical since N is much smaller than
the total number of clients and the total amount of the
addressees of each spam mailing, so your probability to
be the guinea pig is small.  However, if a certain
well-connected client is close to the hose most of the time,
it will suffer unevenly. 

One thing which comes to mind is a post-mortem: run dccm,
without rejection but filing copies separately, then dccproc
it and compare the headers.  For the stuff which was not
counted as "many" originally but proved to be spam later,
make a note that you're closer to the spammer than the
rest (of the world), and...  um, think about the evil in
its many faces.  :)

Alexy Khrabrov :: www.setup.org :: Age Quod Agis

More information about the DCC mailing list

Contact vjs@rhyolite.com by mail or use the form.