On Fri, Jan 09, 2004 at 06:45:57PM -0000, Sabahattin Gucukoglu wrote:
On 9 Jan 2004 at 9:18, Harald Tveit Alvestrand
<harald(_at_)alvestrand(_dot_)no> spoke, thus:
Why doesn't your friend use ETRN to trigger delivery of his queued mail
from his mate whenever he gets online?
He doesn't want his mate getting his mail while he's not available if he
will be available shortly after. The idea is to restrain clients from
passing onto the next MX, and thereby let his mail fall into his own
responsibility when the unavailability of his host is either known to be
temporary or is simply not long enough to justify any resulting policy
differences between hosts.
But the problem is that until vast quanties of people understand your
extended MX proposal, the mail will be received by the backup server
anyway. And given past experience, it takes a long, long time for
support for new features to get propagated to enough of the global
internet infrastructure that it will really solve your friend's
Let me propose another possible solution, which doesn't require making
global infrastructural changes. It will require changes to the SMTP
server on the backup MX machine, but that is something which is much
more likely to be doable.
First of all, on the backup MX server, you create a little daemon that
periodically checks to see if the primary mail server is up. If it is
down for more than a predefined time perioud (a day in your example),
then it sets a flag "this host is down hard".
Secondly, you change the the SMTP server on the backup MX server that
by default, attempts to send mail to your friends domain return a 4xx
soft error, UNLESS the "this host is down hard" flag is set. If that
flag is set, then the mail is accepted, and queued for transmission to
your friend's mail server.
As a side benefit, current spamming software will see the soft
failure, and not bother to retry the mail transmission while your
friend's host mail server is temporarily down.