Re: Fw: Re: MJ2 causing greylisting delays

On Thu, 23 Aug 2007 12:41:32 -0400 (EDT), John L wrote
> > As MJ2 always changes the sender, every new message gets greylisted.
> 
> Right.  That is an indication of misdesigned or broken greylist 
> software. Once the sender retries, any sensible greylisterer 
> whitelists the IP. Only the very primitive greylisters fail to whitelist.

Well, unfortunately not, and it's not used in that way anymore.
As a network manager, I can present several cases why whitelisting the
IP address after one successfull delivery is very naive approach, but
let's take these two:

1) there are many dynamically assigned IP addresses. When you whitelist
an IP address for X days after first delivery and don't care about
sender/recipient anymore, it might get assigned to completely dirrefent
computer the other day and you'll get tons of spam until the whitelist entry
expires

2) yet worse, there are many NATs/PATs hiding several computers behind one
real IP address. One sucessfull delivery would open the door for all those
machines for X days - and depending on actual setup, this might mean hundreds of
computers.

Therefore, an uptodate greylister whitelists [IPaddress,sender,recipient]
for quite a long period - e.g. 30 days - and with every consecutive delivery
checks all those three parameters. This causes no problem to personal email, and
also no problem to most mailinglist software. As said before, all mailinglists
I'm member of have no problem with this setup.

> > and works without any problems with our greylisting setup. Not sure
> > if MJ2 isn't in fact the only package changing the envelope-from for
> > each message.
> 
> Trust me, it's not.  Any mailing list that goes out through qmail, for 
> example, uses VERP per-message bounce addresses.  MJ2's behavior is 
> completely standard conformant and quite normal.

Wait a minute. Qmail's VERP passes through the greylists fine, since it's always
the same:

    owner-listname+user=domain@origin

+user=domain is recipient's email address, i.e. always the same.

However, MJ2 uses message number, which is *different* for every message.
 
Thus, given the arguments above and our experience, MJ2's behaviour is
quite non-standard and I'm sure it can be improved. Afterall, if the other
software can sucessfully work without changing the envelope-sender for
every message and at the same time automatically remove repeatedly boucing
addresses, there *is* an alternative solution. For instance, you can certainly
implement the per-recipient VERP example above.


     With kind regards,

         M.