VERP - Variable Envelope Return Path

[David Relson]

On Thu, 25 Sep 2003 19:08:35 +0200
Matthias Andree <matthias.andree at gmx.de> wrote:

> David Relson <relson at osagesoftware.com> writes:
> 
> > Greetings,
> >
> > Yesterday I discovered my wordlist has a large number of singleton
> > tokens of the form "bogofilter-return-586-relson" and learned that
> > they're called VERPs.  After a google search, and some reading, I'm
> > now an expert --- Not!!!
> >
> > As an experiment I added the following lines to bogofilter's lexer:
> >
> > 	VERP		{TOKEN}-[0-9]+-{TOKEN}={TOKEN}@{TOKEN}
> > 	<INITIAL>{VERP} { skip_to('='); return VERP; }
> >
> > With a little bit of additional code, the following header lines:
> >
> > 	From bogofilter-return-586-relson=example.com at aotto.com  Fri Nov
> > 	1 07:42:39 2002
> 
> Well, the relson=example.com part is YOUR address, so it's invariant
> for that list; the changing part is the message ID, 586 in this case.
> 
> I'd probably go for [0-9a-fA-Z] or even the whole alphanumerical
> range.
> 
> > could yield token "verp:bogofilter-return-X-relson" (as well as
> > other tokens).  This would change the non-recurring single use
> > tokens into a repeated, recognized token that contributes to the
> > spam/ham score.
> 
> > Anybody have a better idea, alternate approach, or reasons
> > for/against making this change?
> 
> Go for it.

The VERP form is {token1}-{token2}-{number}-{token3}={token4}@{token5},
as best I can tell.  How much freedom does the mailing list have in
setting token1 and token2?  Can they have embedded hyphens?  Can they be
all numeric?  If the answers are No and Yes, I'm all set.  Otherwise,
the task is harder.