[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Asrg] pre-rfc thought balloon: ESMTP DATAFIRST



On 2006-06-06 00:21:16 -0000, John Levine wrote:
> >The idea if implemented would ease content-based filtering at the
> >server as messages to multiple recipients would no longer require
> >separate complete transactions for each recipient when sending to
> >servers that implement complex per-recipient acceptance policies.
> 
> I wouldn't bother, for several reasons:
> 
> One is that you're optimizing a rare case and pessimizing an unusual
> case.  I get a lot of mail that I reject at RCPT TO, little mail to
> multiple recipients, and no mail at all to multiple recipients who
> need to do different things with it based on the content.  If one
> thinks it's spam and the other doesn't, that invariably means one of
> them has a misadjusted spam filter.

I agree that the case is rare, but nevertheless I think it is an
important case (I have implemented a plugin for qpsmtpd to allow
per-recipient content-filters - an ugly hack, but it seems to work).

There are cases where it isn't a misadjusted spam-filter (a mail
crossposted to a closed and an open mailing-list comes to mind), and
even where it is, one user's misadjusted spam filter shouldn't prevent a
different user from getting legitimate mail.


> The next is that for reasons unrelated to bandwidth, there are good
> reasons to send list mail one copy at a time.

I agree with that, but messages with lots of recipients are often not
sent through mailing-list software. People create "lists" in their
personal address books and simply send a mail to umpteen recipients.

> The last is that despite the persistent folklore, the amount of
> bandwidth that mail uses is trivial compared to web and P2P, and it is
> a foolish economy to try and reduce it even more.

Last time I looked at our site the smtp traffic was comparable to http
and quite ahead of the p2p protocols I know. I was quite surprised. I'll
have to measure that again soon.

But I agree that bandwidth isn't much of a problem. The problem is that
messages with dozens or hundreds recipients are transferred in a single
transaction and that the SMTP protocol currently doesn't offer any any
method to return per-recipient results after receiving the content.
Current solutions (like my cf_wrapper mentioned above) rely on temporary
failures, which may cause long delays if there are many recipients. 

So I think that some extension should be implemented. Like some others I
would prefer something based on the LMTP model, though. (But maybe
that's just because I already know LMTP)

	hp

-- 
   _  | Peter J. Holzer    | Ich sehe nun ein, dass Computer wenig
|_|_) | Sysadmin WSR       | geeignet sind, um sich was zu merken.
| |   | hjp at hjp.at         |
__/   | http://www.hjp.at/ |	-- Holger Lembke in dan-am

Attachment: pgpQRSSw2f1m5.pgp
Description: PGP signature

_______________________________________________
Asrg mailing list
Asrg at ietf.org
https://www1.ietf.org/mailman/listinfo/asrg