Viktor Dukhovni:
> On Tue, Sep 11, 2012 at 04:26:07PM +0530, DN Singh wrote:
>
> > domain3_destination_rate_delay = 30s
> > domain3_destination_recipient_limit = 1
>
> This combination is unlikely to yield the results you expect. When
> a transport's recipient limit is set to 1, its concurrency is
> measured *per-mailbox*, not per *nexthop*. In this case you get at
This would be a candidate for a helpful message if there were a way
to determine that this is a likely mistake. Perhaps it is safe to
make that assumption when the address class is not "final", i.e.
the "default" or "relay" address class.
There are plenty examples where a per-delivery recipient limit of
1 is desirable, even in combination with a per-recipient rate delay.
Typically these involve delivery to potentially-expensive targets
such as delivery to some local command.
Wietse
