spam.whitelist.rules

[John Rudd]

On Friday, Aug 1, 2003, at 01:40 US/Pacific, Julian Field wrote:
>
> This would give 100% predictable behaviour when messages have multiple
> recipients.
> One possible tweak to the above behaviour might be
> 1a. If all the values for all the recipients are the same, then use
> them.
> But that would be more awkward to implement as the entire ruleset would
> have to be evaluated once to check the results, then gone through
> again for
> the *@domain.com or *@* addresses. Would definitely be slower.
>
> Let me know what you think of either of these possibilities.
>

I think slow would be bad, so I'm against that.

I think the method you give might be interesting, but I think it might
also be interesting to give the option to have the message replicated
for each rcpt address.  So, if the message is going to 3 users and 1
mailing list, mailscanner would make it into 4 messages.  The question
is, would this be a general mailscanner preference, or would it be
something within the ruleset?

In the first case, messages would always get split.

In the second case, the ruleset would have a keyword that says "split
the message if it has multiple recipients".  Or, perhaps, the keyword
would say "If you have multiple recipients, and they end up matching
different rules, then split them", but that probably gets back to the
slow approach (perhaps if you build an array of "rule matches" for that
rule and that message, then you'd make one pass to build the array,
then your second pass of evaluations would only be done against the
array, and you would only split the message if the keyword was in that
rule and the array had more than one element).


What I'm planning to do under CommuniGate Pro is this:

With CGP, there are 3 levels of "rule" processing: "Site Wide" (happens
before any sort of expansion), "Domain" (happens after expansion for
both mailing lists and multiple individual recipients, but before user
rules), "User" (rules created by/for individual users, and controllable
by them, mostly for procmail type things, but could also be used to
spawn external programs).  Domain rules are new to the current version,
so I haven't tried that yet... but I'm planning to move my
"CGP-to-MailScanner" gateway into the domain rules.  That means there
will only ever be 1 recipient to each message that MailScanner sees.
I'll combine that with putting the user's settings into LDAP and
building my custom module around extracting their settings from LDAP.