> I had an idea about this: what about allowing random macros to be checked,
> instead of just this particular one? I think about such a syntax:
> sm_macro "may_be_forged" "{client_resolve}" "FORGED"
> acl blacklist domain /.*\.info/ sm_macro "may_be_forged"
>
> It would bring much more flexibility, as any sendmail setting could be
> used in the ACL. For instance, someone asked for a sendmail bound to
> multiple IP with a different milter-greylist configuration for each IP.
> Using the if_addr macro in the ACL would allow that:
>
> sm_macro "ip1" "{if_addr}" "192.0.2.3"
> sm_macro "ip2" "{if_addr}" "192.0.2.4"
> acl whitelist sm_macro "ip1"
> acl greylist sm_macro "ip2" delay 15m autowhite 3d
>
> Likewise, SMTP AUTH status could be used in the ACL just by checking
> {auth_authen}.
>
> I wonder if going further is of any interest: should we support regexp for
> the macro value? lists of sm_macros? Anyone sees an usage for that? Here is an
> example of macro + list that we could support:
>
> sm_macro "may_be_forged" "{client_resolve}" "FORGED"
> sm_macro "blacklist" "{blacklist}" "BLACK"
> list "bad_macros" sm_macro { "may_be_forged" "blacklist" }
> acl blacklist domain /.*\.info/ list "bad_macros"
>
> Any comment?
1. What is the advantage of giving names to macro conditions rather
than the following syntax? Line length?
acl greylist macro "{client_resolve}" "FORGED"
list "bad_macros" macro "{foobar}" { "foo" "bar" }
2. If we introduce your sm_macro syntax, we need to consider the
sm_macro as a part of ACL rather than config. The sm_macro must be
protected by the ACL lock and may be inconsistent with the config.Message
Re: [milter-greylist] RFC: checking sendmail macros in ACL
2006-08-25 by AIDA Shinra
Attachments
- No local attachments were found for this message.