We implemented a EzMLM like bounce system for our MLM... The VERP (return address) looks like: [EMAIL PROTECTED] Or in your case [EMAIL PROTECTED] We then run the following state machine: If you bounce back 10 in 3 days we will warn you. Note that if this is bounced back we just log it and carry on. If you bounce back 20 in 6 days we will send you a bounce probe. If we get back the bounce probe you WILL be marked as bouncing and no more email will go out.
This is (roughly) how EzMLM works IIRC, but we have tried this system out in the field and it works fairly well. I'd be interested to know if you managed to get meaningful information out of the bounced email. We have tried using a regex to pick up the 4xx and 5xx codes and it seemed to work sometimes. However that code has NOT been tested in the field so I don't know how reliable it is. It works with Qmail. YMMV. I think it would be a cold day in hell if Lotus or MS give you good quality error codes... -- Jason > -----Original Message----- > From: Serge Knystautas [mailto:[EMAIL PROTECTED] > Sent: 31 July 2003 14:26 > To: James Developers List > Subject: Re: More info from bounces > > > Noel J. Bergman wrote: > >>The motivation for this is to recognize soft bounces (mailbox full, > >>mail servers down) from hard bounces > > > > Servers should probably not be sending a bounce, or at > least a non-RFC > > 1894 bounce, for a soft error. VERP seems to be somewhat > predicated > > upon bounces being permanent, at least for that message. To work > > around the fact that the errors may relate to a transient mailbox > > error, it seems that VERP mailers will be tolerant of a configured > > number of bounces, and will then send an "warning message" > to inform > > the user. If the warning bounces, the user is disconnected > from the > > list. That seems to describe the behavior I've seen from > ezmlm, but > > Brian (or perhaps some of our members) would know better. > > Here's an update on my data, and actually makes me think we need to > hurry up and add this. :) > > # of bounces: 1374 > # of James bounces: 541 > # of non-James bounces: 833 > # of non-James bounces w/RFC 1894: 514 > % of non-James bounces w/RFC 1894: 61.7% > > I've only glanced at the spec so don't really know how much work it > would take, but it seems this is much more widely adopted than I had > thought. > > -- > Serge Knystautas > President > Lokitech >> software . strategy . design >> > http://www.lokitech.com p. 301.656.5501 e. > [EMAIL PROTECTED] > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
