Great Circle Associates Majordomo-Users
(June 1994)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: handle bounces automaticall?
From: Alan Millar <amillar @ bolis . sf-bay . org>
Date: Thu, 30 Jun 1994 01:55:42 -0800 (PDT)
To: rouilj @ cs . umb . edu (John P. Rouillard)
Cc: reisert @ wrksys . enet . dec . com, majordomo-users @ GreatCircle . COM
In-reply-to: <199406291438.AA15642@cs.umb.edu> from "John P. Rouillard" at Jun 29, 94 10:38:48 am
Reply-to: Alan Millar <AMillar @ bolis . sf-bay . org>

And verily didst John P. Rouillard spake of these matters:
> Interesting idea. As long as you can find an actual error line for the
> mailer and differentiate it from informational messages with addresses
> (such as those generated by deliver/procmail scripts etc) it seems
> workable. The one thing that the program has to have is a very low
> (non-existent) false positive rate (i.e. it unsubscribes somebody who
> wasn't bouncing mail). 

That's the trick, of course: how to catch enough of them without any
false positives.

I think I have it narrowed in fairly tight.  I think I'm on the 
conservative side enough that variations in bounces are more likely
to be skipped than bounce a good address.

However, I would like to see any message that anyone thinks might
trigger a false positive.

> >the list.  Note that there is a separate listname-autobounce alias
> >for each list which passes the list name to the autobounce program.
> >That way the scope of possibilities is much narrower, and we are
> >more likely to get good results because we can check to make sure
> >what we think is the bouncing address really is a list member.
> 
> One thing I would suggest is looking for an X-listname header or some
> such as a sanity check on determining the list name. This would allow
> it to be run from a .forward file for instance.

I don't know how safe that is.  By running it on the list sender address,
you know you'll only get bounces from that list's postings.  I would
tend to NOT put it on a general address that receives all sorts of messages
because I definitely think that would greatly increase the chances of
a bogus match.  As I said, putting it on the list sender address
will pretty much limit the messages to real bounces from that one list.
 
> You might also want to make configuration options for:
> 
> 	enabling/disabiling the autobounce program
> 		(autobounce = yes/no)
> 	the name of the bounces list, or "none"
> 		(bouncelist = <word>)
> 
> If you want, I can do the config options before you release autobounce.

It also needs a configuration option for the approval address.  This
is where the majordomo results are send back to.  Defaults to
listname-approval.

> >One thing the program needs is a good way of specifying the bounce 
> >message profiles so they can be easily added to.  Right now it
> 
> 	Put them in a file called autobounce_pat.pl (or whatever), and
> 	just require it.

That's what I ended up with.  I called it autobounce.cf

- Alan

----
Alan Millar          E-Mail: amillar@bolis.SF-Bay.org
System Administrator    Web: ftp://ftp.netcom.com/pub/amillar/welcome.html
Abortion stops a beating heart



References:
Indexed By Date Previous: Re: handle bounces automaticall?
From: Alan Millar <amillar@bolis.sf-bay.org>
Next: Netcom and digesting
From: sylviac@netcom.com (Sylvia Caras)
Indexed By Thread Previous: Re: handle bounces automaticall?
From: "John P. Rouillard" <rouilj@cs.umb.edu>
Next: Re: handle bounces automaticall?
From: mmorse@nsf.gov (Michael H. Morse)

Google
 
Search Internet Search www.greatcircle.com