kristin(_dot_)hubner(_at_)Sun(_dot_)COM wrote:
> But in other cases, with more "unfriendly" bounces, rejecting the message at
the SMTP level is more
> desirable. When it's more important to limit load on or protect your own server vs.
being "friendly" to
> the original sender, then "reject" can be more appealing.
>
> I can therefore envision cases of Sieve scripts that might wish to use both.
This isn't specific to just Kristin's comment, but if you're talking
about using sieve in the SMTP transaction, when does this happen?
I suppose it could happen anywhere in the dialogue, but as a practical matter,
any script that uses header tests (and the majority of them do) can only
after the message has been transferred.
Of course this runs into problems when there are multiple recipients.
But remember: The majority of messages have a single recipient. And there's
also LMTP to consider - a sieve implementation operating in the context
of an LMTP server has no problem returning per-recipient failures after
the data.
Ned