Milter-greylist

Just a sanity check on two samples...


# sample one:
racl whitelist from "dude@..."
racl greylist default                  delay 30m autowhite 5d
dacl greylist body /hello world/       delay 10m autowhite 9d

# sample two:
racl greylist from "dude@..."  delay 30m autowhite 5d
racl whitelist default
dacl greylist body /hello world/       delay 10m autowhite 9d


dude@... sends an email with "hello world" in the body.
Sample one delays it for 10 minutes and sample two delays it for 30
minutes (quietly skipping over the dacl greylist entry).

dude@... sends another mail, this time without "hello world"
in the body.  Sample one does not delay him at all while sample two
again delays him thirty minutes total.

bub@... sends an email saying "hello world" in the body.  It
is delayed by 30 minutes in sample one and 10 minutes in sample two.

bub@... sends an email without "hello world" in the body.
Sample one delays him 30 minutes while sample two has no delay.


Right?  Wrong?  Warnings?  Unintended effects?

On Tue, Jul 14, 2009 at 04:35:33PM -0400, Adam Katz wrote:
> dude@... sends an email with "hello world" in the body.
> Sample one delays it for 10 minutes and sample two delays it for 30
> minutes (quietly skipping over the dacl greylist entry).

Yes: because greylisting tempfailed at RCPT stage, DATA-stage ACL
are not evaluated. That one is consistent.

> dude@... sends another mail, this time without "hello world"
> in the body.  Sample one does not delay him at all while sample two
> again delays him thirty minutes total.

You pass the RCPT-stage ACL and then you get a match at DATA-stage. 
DATA-stage ACL (if evaluated) overrides the RCPT-stage ACL result.
Is it desirable? No idea, but at least this is the way it works now.
 
> Right?  Wrong?  Warnings?  Unintended effects?

Well, the problem is that conjunction of RCPT and DATA stage ACL in
milter-greylist has never been fully specified, so you get some
odd results, sometime.

-- 
Emmanuel Dreyfus
manu@...