Re: [Exim] Exim 4 config file: opinions sought

Top Page
Delete this message
Reply to this message
Author: Steve Sargent
Date:  
To: exim-users
Subject: Re: [Exim] Exim 4 config file: opinions sought
I think I would prefer the big revision. It produces a more logical
flow of operations.



>When I added authenticators to Exim, I put their configuration at the
>end of the file, so as not to invalidate existing configs. I am now
>thinking about adding the ACL section (as previously discussed) for Exim
>4. Because Exim 4 has no directors (just routers and transports), its
>configuration file is necessarily going to be incompatible, so there is
>the possibility of putting the ACL section somewhere other than at the
>end. This has made me wonder whether there should be a general revision
>of the order of sections. Here are some possibilities:
>
>1. Minimal: Remove directors, put ACL at the end, leading to
>      . main
>      . transports
>      . routers 
>      . retry
>      . rewrite
>      . authenticators
>      . ACLs
>    At least the sections that control input (rewrite, authenticators, ACL)
>    and those that control output (transports, routers, retry) are together.

>
>2. A slight variation could be to put ACL before rewrite, as it is
>    probably the most important of the input-controlling sections.

>
>3. Big revision: Put input before output, leading to
>      . main
>      . ACLs
>      . authenticators
>      . rewrite
>      . transports
>      . routers
>      . retry

>
>4. Transports originally appeared before routers (and directors in Exim
>    3) so that transport names given in the routers could be checked for
>    existence when the config was read. Subsequently string expansions for
>    transport names were introduced, and these cannot be checked till
>    they are used, although in most configs fixed names are still used
>    (and checked). However, it takes some people a while to understand
>    the order in which the drivers are used, and I wonder if it might help
>    to put routers before transports? (And give up on the read-time
>    check.) I suppose a similar argument suggests that authenticators
>    should come before the ACLs, because authentication happens before
>    the ACLs are consulted.

>
>This is all pretty minor, really. Does anybody have any strong views?
>Any views at all?
>
>--
>Philip Hazel            University of Cambridge Computing Service,
>ph10@???      Cambridge, England. Phone: +44 1223 334714.

>
>
>--
>## List details at http://www.exim.org/mailman/listinfo/exim-users
>Exim details at http://www.exim.org/ ##


--
|--------------------------------------------------------------------------|
| Steve Sargent, Vox +44 020 7775 3220, Fax +44 020 8980 2001              |
| QMW Computing Services,  Mile End Road,  London E1 4NS, UK               |
| Email   : S.L.Sargent@???                                          |
| WWW page: http://www.qmw.ac.uk/~cgaa160/index.html                       |
|                                                      |                   |
|                                         PIPER       _|_                  |
|                                         PA28R  ____/___\____             |
|                                         ___________[=o=]___________      |
|                                         ARROW    e/  o  \e               |

|--------------------------------------------------------------------------|