On Sun, 13 Nov 2005 16:09:44 -0600, Hal Keen <[log in to unmask]> wrote:

>The email response depends on the user being able to send the
>command from the subscribed address; as noted above, not all can do so.

I think that in most cases (the CHANGE command is a known exception), the
From: address used for the OK-Confirm operation need not match the
address-mailed-to by LISTSERV as long as the cookie code does match.

>When a Probe is sent to a nonworking address, the error goes undetected.

You have mentioned something like this before.  This would suggest some
possible problem with your installation that prevents such bounces from
getting back to LISTSERV.  If they do get to LISTSERV the bad addresses will
be auto-deleted (assuming you have sensible autodel settings).

Probe messages (both passive probe and active probe) go out with a modified
Return-Path address:

Return-Path: <owner-listname*bparker**BESTEFFORT*[log in to unmask]>

Note that the address-mailed-to is encoded into the normal return-path so
that even if the bounce report comes from another domain/address, LISTSERV
will know the address it mailed to is the one at fault.

Some mail systems have a problem with the format of this address, even
though it is perfectly "legal" per RFC821 (and RFC 2821).  Sites running
LISTSERV on unix need to use the "sendmail hack" so that all owner-*
addresses are delivered to LISTSERV, even though it does not perfectly match
the usual owner-listname alias in your /etc/aliases file.  Windows sites
normally don't have this probem since both SMTPL.exe and LSMTP.exe (which
receive the mail for LISTSERV) understand this special 'probe format' return
address and know what to do with it.