[Spambayes] SpamBayes questions

[Meyer, Tony]

> I have approx. 6 @getsoft.com I want to filter, and possibly 
> some others. Seeing I only discovered Spambayes yesterday and 
> installed it today, I really don't have much of an 
> understanding how it goes about doing what it does.

:)  A reason things get tricky with multiple users is that you have to
decide whether they are going to share a database (and hence, a common
definition of spam/ham).  This leads to issues about storing separate
databases, or issues about training shared databases via separate
interfaces (so that not all mail is seen by all users).

The multiple users/server question is asked so often that it would be
really great (IMO) if we did have an example of how to do it (just
making arbitrary decisions, and pointing out where they were).  If I
find time and finish the other things I'm meant to be doing ;) then I
might play around with this (it would be time consuming to set up,
though since I would have to create a test mail server on my system,
since I don't have access to one).

Anyway, in the meantime you can see how things go with the local system.

> Typically my email client is setup to delete mail as it 
> downloads it. Is it feasible to have a mode/setting which 
> says don't pass spam on to the email client and also don't 
> delete it from the mail server. If the email client never 
> sees the spam in this special mode then it won't delete it 
> and it will just sit there. Ideally this "mode" could be set 
> by the Web interface. I could still use WebMail to look at 
> the mail back on the server if I really wanted to.

What I could see as an 'easy' option would be this:
* your mail program can connect to pop.example.com:110 and all mail
waiting is retrieved normally (and you probably delete it all from the
server as well)

* your mail program can connect to pop.example.com:111 and only
ham/unsures are downloaded; all spam (and false positives ;) are left on
the server (and you probably, again, delete all the mail you download,
since you have a local copy of it).

* your web browser can connect to pop.example.com:80 and see the web
interface, which lets you correct any misclassifications, review waiting
(spam) messages, and the like.  If you reclassify a message from spam to
ham, then it will be downloaded next time your mail program connects to
:111.

This way you would only need to connect to :111 and occasionally review
via the web ui.  Alternatively, you could normally connect via :110 (and
filter based on the headers, perhaps), and connect to :111 when you are
on a dialup, or a mobile, or whatever.  The mail clients connecting to
:111 would have no idea that there was more mail there (or, perhaps,
there could be an option to download a stripped version of the message
so that you know that there is mail that you can check, or maybe a
single summary message occasionally).

It's not that tricky a project (IMO), if the shared user logistics
mentioned above are resolved.

Actually, thinking about it, Death2Spam (http://www.death2spam.com) does
something along these lines.  I didn't read the docs all that carefully,
so there are probably differences, but it's similar, at least.

> I really appreciate your quick response and willingness to 
> help. Kudos to all those involved in this project.

No worries, and thanks :)

=Tony Meyer