On Thu, 12 Apr 2007, Leszek Dubiel wrote:
> I understand option "unseen" like this:
>
> 1. I start with a "nowak" recipient
> 2. unseen router makes two copies of that addres (because it is unseen):
>
> -- one copy is passed to the next router (*)
> -- second copy is handled with this router (**)
>
> 3. From now on routing is "forked" and addresses are copletely
> independent. Address (*) was not even touched with unseen router.
Yes, that is essentially correct.
> I can see three opportunities:
>
> "header additions don't change address (*)"
> "header additions are deleted from address (*) because router is unseen"
> "header additions are deleted from address (**) if this is passed to
> subsequent routers"
It is documented that what happens is the second of those.
> I would prefer to think like this: "all options of unseen router for
> address (*) are ignored, so header additions don't take place either".
Well, I am afraid that is not what happens. All header additions and
removals that have so far been accumulated during the routing are
carried on with ** but not with *. In effect, * starts out "fresh".
> I would prefer to think: "header addtions are not applied to nowak,
> because this is just passed to the next router" instead of "header
> additions are deleted from address nowak, because they were added and
> router is unseen".
Again, sorry, it isn't like that. All previous header processing is
assumed to be relevant to the copy that is "forked off" at this point.
The copy that continues is "fresh". There were good reasons for this
when it was done like that, but I cannot recall exactly what they were.
> Yes. When I have two copies of "kowalski" -- one with added headers, and
> another without them, then Exim will delete one of these and I don't
> know wich one (this depends of deletion algorithm). So option
> repeat_use=false must be set.
Exactly. I have added some words to the next edition of the manual to
try to clarify this a bit.
--
Philip Hazel University of Cambridge Computing Service
Get the Exim 4 book: http://www.uit.co.uk/exim-book
