Kerry said:
> I have noticed many emails arrive to me with gibberish in the subject
> header. An option to search a dictionary for correct words would be nice.
It
> may be difficult to implement, I don't know. I had an idea the subject
> header could be searched, and if more then a certain percentage of words
> were not found as valid, the letter could be trashed.
That's one of several ideas for the "next" big enhancement to TF. I'm going
to send out a survey asking for input once I finish the current redesign of
the internals.
David Kosik said:
> nice idea, but how about non-english speaking countries?
> like mine eg. ;) - czech rep.
No problem, add your words to the dictionary. Such a filter will never work
with a one-size-fits-all dictionary. I'll probably distribute a small
English dictionary with the filter, but I have no illusions that it would
work well for anyone out of the box.
That shouldn't be a problem; there will be a tool included with it that will
allow you to analyze mistakenly blocked messages and add the words in them
to the dictionary.
> bayesian filter does the job IMHO better.
Not on the mail server. They're really only useful on the client (unless
your mail server serves only a single client or very closely related
clients). They also have major overhead problems given that many spammers
fill their messages with gibberish in order to overload Bayesian filters.
Plus, you have to score *all* messages going through the server (based on
their *final* disposition), which is a big overhead adder for users passing
50K messages a day.
In any case, TF is about using carefully targetted filters so that you can
delete messages pointing at a phony pharmacy without worrying that you'll
miss mail on this list that happens to mention the URL in text. Scattershot
filters like Bayesian ones don't have that property.
(I read an article last year that said that the Bayesian filter equations
are numerically unstable. The net effect is that such filters work more
because of the techniques for deciding what to include and what to ignore
(the "art" part of them) than the science of the probability equations. That
implies that they work (if they work at all) because of the skill of the
programmer and because of the stupidity of most spammers, not because of any
real edge over other types of filters.)
Randy.
This is the discussion list for the IMS Free email server software.
To unsubscribe send mailto:[EMAIL PROTECTED]
Delivered by Rockliffe MailSite
http://www.rockliffe.com/mailsite
Rock Solid Software (tm)
