On Thu, 15 Apr 2004, Benjamin Ritcey wrote:
| Well, I've never used messagewall, but I'm pretty sure I know _why_ it
| does that (although I'm not sure it's RFC incompliant); more to the
| point, exiscan may not be any better for your needs than messagewall.
|
| Once the DATA stage of the SMTP conversation starts, the server can
| either accept or reject the message - _for all addresses listed in the
| RCPT stage_; there's no way for the protocol to handle "OK for user1,
| REJECT for user2", etc.
|
| So that means you either need to have the same rules apply to all users
| (e.g., reject all spam with a score greater than 10), or you need to
| force the sender to send to only one user as a time, using the method
| that messagewall does. This applies equally to exiscan - see
| http://duncanthrax.net/exiscan-acl/exiscan-acl-examples.txt, section 6,
| "Having multiple content scanning profiles for several users or
| domains." (that said, you can still generate bounces later in the
| process, but that defeats the loveliness of SMTP-scanning).
You can do much better with Exim - defer at RCPT time, but only where
recipients have differing spam preferences.
Here, we offer users a choice of 2 profiles, so each message is received
in *at most* 2 batches. Whereas with messagewall, every message is
received in N batches, where N is number of recipients.
See Alan's posts describing this:
http://www.exim.org/pipermail/exim-users/Week-of-Mon-20031006/061084.html
http://www.exim.org/pipermail/exim-users/Week-of-Mon-20031006/061151.html
Chris
--
Chris Edwards, Glasgow University Computing Service
