From: "Glenn Meadows" 

The next release of Polarbar Mailer has  Bayesian filters added to it.
Others on the testing team have been working with them for several weeks
now, and say they're very effective.

--
Glenn M.


"Tony Williams"  wrote in message
news:3dfbe15f{at}w3.nls.net...
> Geo. wrote:
> > "Tony Williams"  wrote in message
> > news:3dfb8a09$1{at}w3.nls.net...
> >
> >
> >>Exactly. My preference would be for a two-level system, with automatic
> >>tagging of emails which look like spam (we use this at work and it's
> >>surprisingly accurate), followed by blocking if the spam doesn't stop.
> >
> >
> > Tagging emails is a great test of the filters abilities but it's not
much of
> > a solution to anything. The problem is there is a limited amount of disk
> > space allocated to each mailbox and when you are talking about a mail
server
> > of any size the amount of disk sucked up by spam is a staggering amount.
At
> > a guess, for every 10,000 users (average folk some careful with their
> > address and some not) receive about 100+mb of spam a day.
> >
> > Tagging emails does nothing to conserve this space and provides no
deterrent
> > for spammers or spam friendly ISP's.
>
> Depends on your perspective. As an email user I can set up a filter
> which sends all messages starting with the string "[SPAM]"
to a separate
> folder and either ignore them or send complaints. If you run the
> receiving servers it's true that tagging won't cut down the load but you
> can use it as a sort of probation instead of going straight to blocking.
>
> As always: your server, your rules.
>
> > A lot of people say you can't pass laws to control spam because it's a
world
> > wide network (I agree with this assesment). Others say you can't filter
spam
> > because spam can be made to look like valid email (I agree with this
> > assesment as well). That leaves one solution, it's up to all netizens to
> > stop communicating with sections of the internet that allow spamming.
> > Realistically that's the only solution I see that has a possibility of
> > limited success.
>
> Filters which use Bayesian statistical techniques are surprisingly good
> at identifying spam (no, I don't understand the math very well).
> Blacklists are the ultimate weapon though, and I wish there were some
> sort of official blacklist rather than the multiple private ones we have
> at the moment. Probably won't happen.
>
> --
> Tony
>

--- BBBS/NT v4.01 Flag-4