Splitting based on body content

Splitting based on body content

[Alan Shutko]

I want to autosend all of the mail relating to help desk calls at
work to their own folders.  Easy enough to do with fancy
splitting if people put the number in the header.  But they
don't, and our pathetic help desk system doesn't either.

So, I thought I'd write a function to do it from a ":" rule.
Problem is, I have no idea how to get the body of the article to
search.  Any clues?  It's not copied into the *nntpd* buffer
where you have access to the headers, and while I'm willing to
check another buffer, it seems that the buffer to look for is
different if you're respooling or getting incoming mail.  At that
point, I was way too confused and gave up trying to sort out the
code.

-- 
Alan Shutko <ats@acm.org> - In a variety of flavors!
92 days, 19 hours, 3 minutes, 24 seconds till we run away.
Things equal to nothing else are equal to each other.

Re: Splitting based on body content

[François Pinard]

[Alan Shutko]

> So, I thought I'd write a function to do it from a ":" rule.  Problem is,
> I have no idea how to get the body of the article to search.  Any clues?

I've been much tempted to do something like this, for SPAM filtering, but
feel a bit torn out.  Doing from Gnus would be comfortable, as it offers
a lot of flexibility and configurability, but it would be really slow.
Doing it through `procmail' is faster and a little less apparent for the
user, since done off-line with reading, but the last time I tried it,
receiving a lot of messages at once was burning a notable part of the CPU,
for a very notable time.  I'm thinking about writing a Flex filter knowing
many rules, this should be as speedy as anyone could get, but we are rather
far from Gnus.

-- 
François Pinard   http://www.iro.umontreal.ca/~pinard

Re: Splitting based on body content

[Alan Shutko]

François Pinard <pinard@iro.umontreal.ca> writes:

> I've been much tempted to do something like this, for SPAM filtering, but
> feel a bit torn out.  Doing from Gnus would be comfortable, as it offers
> a lot of flexibility and configurability, but it would be really slow.

That's why I have a PIII 700.  Personally, I'm not concerned with how
long it takes, because it takes much longer for me to refile these
things by hand.  Also, the way I have my fancy-splitting set up, the
body check is only going to happen for a handful of messages a
day... probably about 1-5% of my daily mail input.

I could do it in procmail, piping it to some script which found the
hdc number and created the appropriate spool file, but it's so easy to
make it create a new group in Gnus.  (Well, for subjects.  It would be
for bodies too.)  But I don't seem to recall that Gnus autosubscribed
groups it found through the procmail spool.

Now, spam filtering should probably be done somewhere else, since you
don't care about autocreating groups and you definately want to scan
all messages for spam, so you're going to see a much bigger hit.

-- 
Alan Shutko <ats@acm.org> - In a variety of flavors!
88 days, 21 hours, 54 minutes, 28 seconds till we run away.
She's genuinely bogus.