question on multiple wordlists

[Eric Seppanen]

On Fri, Oct 11, 2002 at 01:02:42PM -0400, Gyepi SAM wrote:
> On Fri, Oct 11, 2002 at 10:51:31AM -0500, Eric Seppanen wrote:
> > The goals are:
> [snip]
> > - allow use of plaintext lists (possibly requiring conversion to db 
> > format) for whitelisting, blacklisting, ignore-listing.
> 
> I can see how ignore-listing would work, but I am not certain
> that the current framework would support (white|black)listing without major changes
> since that kind of functionality requires that we search for the presence of particular
> tokens [1] which then force the decision one way or the other. This would mean, for instance,
> that bogofilter() and register_words() cannot (however indirectly) drive the lexer since we need more
> contextual information than is currently available within those functions. 
> 
> I wonder if we should not leave (white|black)listing to other tools which are better at it.
> Especially given that the markers for xlisting are usually quite grepable and require some
> surrounding context.

I disagree, I think it's quite simple to do whitelisting and blacklisting, 
but since I don't actually have code to do it yet, I won't bother arguing 
the point.  I think it's likely to be expected by users that they should 
be able to have control over what bogofilter does.

I'll keep this in mind, if/when I get to the point where white/black 
listing is possible.

> [1]. I imagine that *listing files would contains lines like:
> 
>  From:me at example.com
>  To:postmaster at example.com
>  Subject: spammy sounding subject

I think anyone using lines like that would be misunderstanding the purpose 
of bogofilter.  I don't mean that white/black listing should be there for 
normal spam filtering.  Perhaps you're right- it might lead people to 
misuse bogofilter, and it might be a bad idea just because of that.  

I was imagining, say, a sysadmin that wants to make sure that his users 
don't start blackholing mail from him.  So he adds himself to a 
system-wide whitelist.

Here's an interesting somewhat-related question: let's say you're a 
sysadmin and you want to install bogofilter for all users that want to use 
it.  However, you fear that the idiots will start subscribing to mailing 
lists, then marking them as spam instead of unsubscribing.  Is there any 
possible way for the sysadmin to try to make such abuse of bogofilter 
harder/detectable?

For summay digest subscription: bogofilter-digest-subscribe at aotto.com