Log in

No account? Create an account



« previous entry | next entry »
17th Feb 2006 | 14:01

One of the problems with SMTP is that there is only one server reply to the message data, regardless of the number of recipients. This means that you can't apply per-recipient message data policies (such as different SpamAssassin score thresholds) without ugly hacks.

CMU Cyrus uses a special protocol for local delivery called LMTP, the local mail transport protocol. It is the same as SMTP except that the initial client command is LHLO instead of EHLO, and the server gives multiple replies to the message data, one per recipient. The reason Cyrus wants this functionality is because of quota checks: it wants to be able to deliver the message to the recipients that are below quota, but defer delivery to other recipients so that retries are all handled by the MTA.

But, as I explained in the first paragraph, this kind of feature would be useful for inter-domain mail transer as well as mail delivery. There's no particular reason that you couldn't use LMTP as it is in this context, and fall back to ESMTP if the LHLO command is rejected. But this is gratuitously ignoring SMTP's extension architecture, and apart from being ugly, it costs an extra round-trip at the start of a connection.

So I have been thinking of knocking up a quick I-D to re-cast LMTP as a proper SMTP extension and advocate its use everywhere. But, wonderful lazyweb! it has already been done: http://www.courier-mta.org/draft-varshavchik-exdata-smtpext.txt (though I would have done some of the details differently).

Of course this is of dubious utility if the spammers don't also use it...

| Leave a comment |

Comments {3}


from: anonymous
date: 18th Feb 2006 19:21 (UTC)

Are you selling your domain-name? dotat.at? :)

Reply | Thread

Tony Finch

Re: Domain

from: fanf
date: 19th Feb 2006 21:12 (UTC)

Don't ask a stupid question if you aren't prepared to make a stupidly large offer.

Reply | Parent | Thread

If the spammers don't use it...

from: anonymous
date: 15th Mar 2006 12:47 (UTC)

...then that's a few fractions of a point on their "should we greylist?" score, or at least it can be. You could then greylist messages that appear to be junk from at least one recipient's perspective; you get not only the normal benefits of greylisting but also a slight incentive to use EXDATA and / or to reformat your messages.

This is much more complex, of course, but changes to SMTP that have all-or-nothing incentives don't ever seem to take off.

Reply | Thread