Re: [Exim] Autoreply which sends no message

On 30-Mar-01 at 16:13:30 Philip Hazel wrote:
> On Fri, 30 Mar 2001, John Horne wrote:
>> We don't run NIS of any sort so there are no file servers which could be
>> used.
>
> I presume you mean "NFS" when you say "NIS"? If not, I'm very confused!
>
Whoops! :-) I meant NFS, but we don't run NIS or NFS. Sorry.

> I think we are talking at cross purposes here. There won't be any
> "nobody" address involved in what I (thought I) suggested. My idea was
> that the slave should do normal delivery, and also send the message to
> the master. Always. So on the slave you might have a director like this
> as the first director:
>
> xxxx:
> driver = smartuser
> headers_add = X-Something
> new_address = $local_part@$domain : $local_part@MASTER.$domain
> new_director = the-name-of-the-next-director
>
> with some special routing that sends MASTER.$domain to the master
> server.
>
Yup, we can do that.

> Then, on the master server, you rewrite MASTER.$domain to $domain,
>
Yup, we can do that as well.

> let the address run through the directors that do the vacation
> stuff - you should be processing a message that is identical (apart from
> an extra received: header) to one that comes directly - but before your
> final delivery director, you put
>
> xxxx:
> driver = smartuser
> condition = ${if eq {$h_X-Something:}{}{no}{yes}}
> new_address = :blackhole:
>
> in order to bypass the delivery.
>
Initially I tried a smartuser director with :blackhole:, but with a
transport specified as well. Sounds convoluted, which it is, and I was not
too surprised that you couldn't do that :-)

With your solution above though, doesn't it mean that the original sender
will get 2 notifications that the recipient is on vacation if the message
originally arrives on the slave system? One from the slave system itself
(via the autoreply transport) and one from the master? I can see that the
above director (on the master) will prevent this from happening with a local
sender (and hence running through the directors), but if someone from the
Internet mails one of our vacation users then the 'I'm away on holiday'
message will have to be sent back via a router. The smartuser director will
not catch it, and I know of no way of telling a router to 'blackhole' a
message. Perhaps I'm missing something? You mention that the 'nobody'
address is not now involved, so I assume the autoreply transport will send
the "I'm on holiday" message back to the original sender.

> Of course, this puts extra work on your master. If you can arrange for
> the slave to know when not to bother (the user is not on vacation) you
> can save some work.
>
This already occurs, so not a problem.

John.