Philip Hazel wrote: > On Tue, 4 Jul 2006, Marc Perkel wrote:
>
> > It would be nice if the Exim syntax were consistent between the ACLs and
> > the transport/routers
> >
> > For example - ACLs allow multiple condition lines, routers don't. ACLs
> > allow !condition - routers don't. Exim would be far easier to work with
> > if the router and ACL commands were more consistent.
> >
> > Not trying to complain - just making a suggestion for improvement.
>
> It's partly history (routers have been there since the start of Exim;
> ACLs were not there until Exim 4), and partly because
>
> (a) In a router, "condition", is an option. Options have one value;
> therefore you can set it only once. All the options in a router are set
> (though not necessarily evaluated) when the config is read. They may be
> given in (almost) any order.
>
> (b) An ACL is more like a script. The order in which things are listed
> matters, because they are obeyed in order. That is why you can have more
> than one of anything.
>
> Now, we could extend some options in routers to be vectors rather than
> scalars, and allow multiple occurrences to set multiple values, but for
> many router options (e.g. "driver"), it would not make sense. It would
> also introduce some notion of order, which would be a new thing. I am
> not convinced that it would actually reduce the diffences very much, and
> it could be equally confusing.
>
> One would have to change the notion of routers to be more script-like -
> which is a large, very incompatible change, to achieve anything like
> consistency.
I had a thought some time ago about this. Instead of going through all that
work, why not have a generic option called acl which calls an ACL to
determine whether to run or not. It could be used similar to condition but
use the acl system. accept would mean this router handles the message.
I have not thought the whole thing through so I don't know what drawbacks
would be.
--
Lab tests show that use of micro$oft causes cancer in lab animals
Got Gas???
