Re: An approach for reducing mail loss with greylisting

Gaal Yahas wrote:
> In discussing greylisting with friends, one of them who was worried
> about possible mail loss because of sucky sender MTAs proposed the
> following:
>
>    How about failing the transaction at the very last moment, after
>    we get the whole message (assuming there's only one, which is by
>    far the most common case)? Then, if after X hours we don't see the
>    same UID again, we *do* deliver the refused message, but mark it as
>    X-Graylist-Status: Abandoned and leave it to procmail/spamassassn/MUA
>    to decide the implications.
>
>    That should work nicely, nay? Greatly diminished chances of mail loss,
>    plus a chance to autotrain your spam filters, and you still get the
>    benefit of getting yourself off spammers' lists.

I don't think it's a fundamentally bad idea.

However one problem you have though is that you still have to consume 
incoming bandwidth, and local disk space, with the full text of messages 
which might otherwise be greylisted away if they were rejected earlier 
in the SMTP conversation.

If greylisting at RCPT time would be rejecting 95% of your incoming 
traffic, then changing the regime to the one you describe would tend to 
increase your network usage and local disk usage by a factor of 20.

So you'd probably want to take that into consideration before deciding 
if it was right for you.

Bill
-- 
This message is subject to Tao's e-mail terms ( http://e.tao-group.com/1 )