Hi!
I have now another solution, that is better, I think:
There are two hosts, einstein and raman, serving the users home on both
side of the WaveLAN. einstein is the main MX, raman the MX for the other
side. ramanuser is a user behind the WaveLAN. On einstein the files
~alias/.qmail-ramanuser and ~alias/.qmail-ramanuser-default exist and contain
"|/usr/bin/forward $[EMAIL PROTECTED]"
The homedir of ramanuser is NFS-mounted on einstein from raman. The information
is deployed via NIS.
Under normal conditions einstein retrieves all mail and delivers them to the
users homedir, for some user via NFS. Now if the NFS is not accessible
because of an WaveLAN-outage, the ~alias/.qmail-ramanuser* files are used
and the mail should be forwarded to raman via SMTP. And because raman is
not reachable, the mail gets queued, generated by a temporary failure.
Without that qmail-local would generate a permanent failure, resulting
in a bounce message.
The setup on raman is similar, it provides local mail delivery on the other
side of the WaveLAN.
Any comments?
Greetings
--
Robert Sander www.gurubert.de
