Re: [Exim] Multiple recipients and whitelist checks

On Fri, 2004-07-16 at 12:01 +0100, Tony Finch wrote:
> When I have time to implement SMTP-time SpamAssassin checks here I'd like
> to allow my users to specify a threshold, which will probably split them
> into about 15 sets (for no threshold and the integers between 5 and 20).
> This implies quite a lot of deliveries for a message with lots of
> recipients, and possibly delays of hours.

I've been toying with this kind of idea too. But why split the message
unless the data payload actually causes a different _result_ for
different recipients?

Consider the case where you have two recipients, whose thresholds are 5
and 10.

If you receive a mail with SA score < 5, or with SA score > 10, you can
either accept or reject it for _both_ recipients. It's only the case
where the score of the incoming mail is _between_ the recipients'
thresholds that causes a problem.

With thresholds that don't vary _wildly_, that case should be fairly
uncommon, surely? I wonder if you could defer the message after DATA
only if that happens, then split the message at RCPT time when it's
retried?

So in the common case, you accept messages with multiple deliveries just
as you did before. It's only when the SA score of the offered mail
_actually_ makes it ambiguous that you defer it and make the sender
submit for each recipient individually.

Btw, how many MTAs are there out there which will drop a message
delivery attempt _entirely_ if they receive 4xx to RCPT TO:, rather than
continuing and delivering the message to the recipient(s) for whom RCPT
TO: _was_ accepted?

--
dwmw2