[9fans] (no subject)

[9fans] (no subject)

[David Presotto]

I just went through a thousand or so 9fans archived messages looking for spam.
About 70% were html or had html attachments.  A few valid messages also
had html attachments.  Just rejecting html would be helpful but
still let a lot through.

We could enter the arms race and throw an automated filter
or go moderated (not really moderated but people filtered).
The latter is the only one likely to be 100% effective, but
a Bayesian filter might be good enough.  Even the flames on
the net are stylized enough to be recognizable as real
9fans mail.  Of course even the Bayesian one needs a
moderator(s) that will reclassify any misclassified stuff
so that the filter will keep up with a changing world
so some people will have to do extra work.
However, it won't insert people delays in normal delivery.

Re: [9fans] (no subject)

[Sam]

What list software are we using?  Is it smart
enough to be told to only bother filtering
new messages, ie, let all messages pertaining
to an existing (recent) thread through?  If
so, we could be a little more strict on the
filtering.

Cheers,

Sam

On Mon, 24 Feb 2003, David Presotto wrote:

> I just went through a thousand or so 9fans archived messages looking for spam.
> About 70% were html or had html attachments.  A few valid messages also
> had html attachments.  Just rejecting html would be helpful but
> still let a lot through.
>
> We could enter the arms race and throw an automated filter
> or go moderated (not really moderated but people filtered).
> The latter is the only one likely to be 100% effective, but
> a Bayesian filter might be good enough.  Even the flames on
> the net are stylized enough to be recognizable as real
> 9fans mail.  Of course even the Bayesian one needs a
> moderator(s) that will reclassify any misclassified stuff
> so that the filter will keep up with a changing world
> so some people will have to do extra work.
> However, it won't insert people delays in normal delivery.
>

[9fans] spam filtering

[Russ Cox]

As Scott said yesterday, the list uses Mailman.
http://www.gnu.org/software/mailman/mailman.html
As he also said, he can't do anything that isn't a stock
Mailman option.  If you want to suggest something,
please look in the documentation first.

We've all gotten more metaspam in the last two days
from this list than we have spam.  Please stop.

Russ

Re: [9fans] spam filtering

[Sam]

> As Scott said yesterday, the list uses Mailman.
> http://www.gnu.org/software/mailman/mailman.html
> As he also said, he can't do anything that isn't a stock
> Mailman option.  If you want to suggest something,
> please look in the documentation first.
>

Actually, what he said was anything that wasn't a stock
option "would require actual work."  I took that to
mean we're looking for a least effort solution to the
problem, not that we're handcuffed by mailman.  I
think it's already been stated that the stock features
aren't sufficient.  Before I go hacking, which version
are we using?

Sam

Re: [9fans] spam filtering

[Sam]

Scott,

I downloaded 2.0.13 since 2.0.11 wasn't
available.  The file
	$MAILMAN/Mailman/Handlers/SpamDetect.py
where MAILMAN is your install directory
is a hard coded spam filter file.  Adding
an RE for html or anything other combination
of incantations seems easy.  The current code
does an OR (if any RE matches, fail).  I would
suggest adding an AND mechanism such that
any message lacking 'RE:' in the subject
header gets sent through a rigorous set
of RE filters (html, nigerian, etc).  I can
come up with something if you don't want to.

Btw, that's some nasty Python.  I hope it's
not representative of the language.

Cheers,

Sam


On Mon, 24 Feb 2003, Sam wrote:

>
> > As Scott said yesterday, the list uses Mailman.
> > http://www.gnu.org/software/mailman/mailman.html
> > As he also said, he can't do anything that isn't a stock
> > Mailman option.  If you want to suggest something,
> > please look in the documentation first.
> >
>
> Actually, what he said was anything that wasn't a stock
> option "would require actual work."  I took that to
> mean we're looking for a least effort solution to the
> problem, not that we're handcuffed by mailman.  I
> think it's already been stated that the stock features
> aren't sufficient.  Before I go hacking, which version
> are we using?
>
> Sam
>
>
>