Autor: W B Hacker Data: A: exim users Assumpte: Re: [exim] $interface_address,
was: Paging Marc Perkel: rDNS on junkemailfilter.com checking
host(s)
Andrew - Supernews wrote: >>>>>> "W" == W B Hacker <wbh@???> writes:
>
> >> The value is indeed available in routers/transports. It seemed
> >> that it might be useful there...
>
> W> It most assuredly is - IP-based virtual domains can be given the
> W> correct IP and matching HELO so the far-end lookup will
> W> match. (assuming the DNS is in order, of course..)
>
> Here is a pop quiz:
>
> Suppose you use the interface = option in a transport to set the
> outgoing interface based on the incoming interface or the source
> domain or something similar.
>
> Under what conditions will exim ignore that, and deliver outgoing
> messages from the "wrong" interface?
>
> And is this a bug?
> Well - to avoid chasing more than one critter at once, let's specify that the
desired IP is hard-coded.
I'll then take a SWAG and say 'if specified interface is in-use by another
process, but another one is available'.
That on the basis that Exim has no problem sharing when there is only one IP for
all work 'bothway', but also that 'unless otherwise directed' (by interface = )
will leave the choice of which of several interfaces to the OS.
Disregarding the transport option directive would be a bug, IMNSHO - IF that is
the case, if only becasue *anything* unpredictable (/dev/random excepted..)is a bug.
Setting to 'queue_only' and not overly agressive timing could probably prevent
it until a fix is in hand.
