Lisp to Python translation criticism?

damien morton morton at dennisinter.com
Tue Aug 20 21:13:24 CEST 2002


I guess you get a better class of spam than I do :) You certainly get
a lot more of it than I do.

I neglected to mention that I have several accounts, the most prolific
of which of which I apply a more stringent rule to. This rule rejects
emails with the word "unsubscribe" in them.

You omitted the test for <your-email-address> in the body of the spam.

Also, the recipients test is for email sent solely to you, not for
mail obviously addressed to you. That is, your email address is the
only recipient. Anything sent to to multiple or unknown recipients
(including to you), from someone not in your whitelist, is rejected.
Anything sent only to you is accepted.

Id be interested to see what your results look like once you implement
the rules properly (assuming you havent already).


Erik Max Francis <max at alcyone.com> wrote in message news:<3D61F53F.EEA15541 at alcyone.com>...
> damien morton wrote:
> 
> > If I am not the only recipient, or the list of recipients is unknown,
> > its spam.
> > 
> > If the body of an incoming email contains <my-email-address> or the
> > words "click here", its spam.
> > 
> > Thats it. I get about 40 spam a day. Perhaps once a week, the filter
> > fails and one gets through. Ive had one false positive in the last 6
> > months.
> 
> It's true that a set of simple rules will get "most" spam.  A handful of
> simple rules, even rules which don't address content, will catch 80-90%
> of spam.  But when you get hundreds of spam emails a day, "most" doesn't
> cut it.
> 
> Having had to put together a very detailed rules-based spam filter over
> the years (and having recently migrated to one which uses Python), I
> quite frankly severly doubt your statistics.  You're suggesting that
> only one spam message in three hundred doesn't meet the above stated
> criteria; I highly doubt that.
> 
> Yesterday, I got 405 emails, of which 391 were spam.  Of those 391, 190
> were not obviously addressed to me (no valid email address of mine
> appeared in any of the usual headers).  Also, of the 391, only 124 had
> one of the phrases "click here," "click reply," or "click now."  233 had
> one or both of these characteristics.  That's a long way from 391.



More information about the Python-list mailing list