Julian Field mailscanner at ecs.soton.ac.uk
Tue Feb 24 12:23:39 GMT 2004

At 11:36 24/02/2004, you wrote:
>Has anyone had look at DSpam?
>See http://www.nuclearelephant.com/projects/dspam
>The author seems to write off SpamAssassin as a filter totally (see
>http://www.nuclearelephant.com/projects/dspam/faq.html#1.7 ) and perl
>implementations in general
>any thoughts on that?
>Regards, Tony

It's very much an aggressive sales pitch. His example of his largest site
at 125,000 mailboxes is not very big at all. I have sites with over 100
times that amount. As for his "peaks at 99.984% accuracy", that's just like
an advertisement telling you that you can get "up to 30% off in the sale".
If he chose his sample right, he should be able to say that it peaks at
100% accuracy.

Sounds like a bunch of Bayes-based or similar approaches. If he is relying
on 1 tool like this, he is doomed to failure as the spammers work round his
filters. The more popular he gets, the faster his approach will die.
SpamAssassin succeeds through its use of so many different approaches
blended into 1 system. It is easy to fool one or two of them at once, but
very hard to fool all of them at the same time.

SpamAssassin Myth 2: you just set up a spam and notspam address just as
described countless times on this mailing list.
SpamAssassin Myth 3: only true if you use the "spamassassin" script, which
almost no-one does. MailScanner certainly doesn't suffer this problem.

Oh, and while we're at it, Perl is not an interpreted language, it's a
just-in-time compiled language. It just looks like an interpreted language.

This just reads like a bolshy sales talk. "I'm wonderful and everyone else
is c**p, and I'm going to keep telling you!". I fully expect there are some
good components in there, and I expect the SpamAssassin guys have looked at
it and made their own judgement on whether there are useful ideas.

Slagging off all the opposition isn't actually a very good way of
convincing people of your argument. Some of what he says is true, but
certainly not all of it. And since I can quickly see several mistakes in
his information, how much else is made up?
Julian Field
MailScanner thanks transtec Computers for their support

PGP footprint: EE81 D763 3DB0 0BFD E1DC 7222 11F6 5947 1415 B654

More information about the MailScanner mailing list