Re: [Nmh-workers] [patch] filtering support for inc

[Jon Steinhart]

Ralph Corderoy writes:
> Hi Ken,
> 
> > If the goal is to filter messages at "inc" time, then I'd guess that
> > the best thing to do would be to add a new interface that allowed you
> > to tell "inc" where you wanted the message to end up.
> 
> Sounds complex.  How about giving inc(1) a [-sequence foo]... that added
> the incorporated emails to the sequences.  The user's script could then
> cook up a unique sequence name, run inc, then pick(1) and mark(1) their
> way through that sequence doing what they liked.
> 
> The unseen-sequence profile entry is a more limited form of this, and
> could be used at a push with a tailored profile for inc's run, if
> needed.
> 
> -- 
> Cheers, Ralph.

I'm curious, what is trying to be accomplished by inc filtering?

Again, I do this using a sendmail milter because I don't want to see the
filtering in action.  But, I'm using it just to partition things into
real mail and spam.

I have a crude but very effective spam filter whose main rule is that if
a message is not from someone in my whitelist and is not text/plain then
it's spam.  Anyone sending me 100x the number of bytes because it's a
multipart message that says "hello world" is unlikely to be someone with
whom I'm working.  I have a few extra rules, such as "if it's to multiple
recipients on my machine and any of those recipients are not valid addresses
then it's spam".

I have a dummy "spam" user, and the milter puts the spam in the spam user's
queue instead of mine.  But, because this is my system I'm in the mail group
and can separately inc the spam into its own folder so that I can scan it
once in a while.

I could do this by filtering at inc time, but then I'd see all of the spam
going by.  That wouldn't work for me since I get at least 100x as much spam
as legitimate email.  I suppose that there a filtering mechanism could have
a "don't show stuff being sent to this folder" setting and then it would
work for me and be a better solution than the milter.

BTW, something that I have contemplated adding is a time element which would
say that messages from different senders with the same content are spam.  For
example, I have recently seen a large number of message with bodies of "hello,
test".  Not sure what they're probing, but since they each come from different
senders they're clearly not legit.

Anyway, just curious as to the use cases from people wanting the feature.

Jon