Automatically moving marked mails?

[Ian Eiloart]

--On 9 July 2009 09:54:31 -0400 Adam Tauno Williams <adam at morrison-ind.com> 
wrote:

>>> Ian Eiloart wrote, at 07/09/2009 05:39 AM:
>>>> Except that the sieve server ought to be on the border MTA, so that the
>>>> user can tell the server to reject the message at SMTP time.
>>> That's not feasible for mail with multiple recipients.
>> It is if your rule is to reject all email from a specific sender.
>
> No, because the MTA either accepts or rejects a message [in
> connection].

Not true. The MTA can decide *per recipient* whether to accept mail from a 
specific sender. It's true that the MTA hasn't seen the message content at 
this point, but it does have enough information to determine - for example 
- whether the sender is a member of a mailing list, or is on a recipient's 
blacklist or whitelist. We do a lot of that.

Exim, for example, can do this in its ACLs. It doesn't have a built in 
SIEVE facility at this stage, but certainly can consult recipient specific 
blacklists. It can even be built with a perl interpreter, so you could 
check for sender conditions in SEIVE scripts.

> If a message is sent to userX and userY and userX has
> SIEVE set to reject the message and userY does not then the MTA has to
> receive the message in order to deliver it to userY.  And the MTA
> would have to check every recipient's SIEVE script.

Sure, there are some content dependent conditions that could not be tested 
at this stage. In principle, they could be ignored for the moment. There 
probably aren't any SIEVE implementations that do what I suggest, and the 
implementations wouldn't be simple, but there's no principled reason that 
it shouldn't.

> Then what about
> delivery to an alias that expands to multiple users?
> Mail delivery
> just isn't that simple.


-- 
Ian Eiloart
IT Services, University of Sussex
01273-873148 x3148
For new support requests, see http://www.sussex.ac.uk/its/help/