[Mimedefang] Strange duplicate mail

[Jonas Eckerman]

On Wed, 15 Oct 2003 17:07:20 -0500, James Miller wrote:

> After the '<enter> . <enter>' it's taking 60-70 seconds to accept
> the message.

That seems rather slow for a short test message. Is the server low on 
CPU or RAM?

> message.  I think some of the servers sending us mail are closing
> the connection before we acknowledge receipt -- and later resends

If you know wich servers does this, you might want to send the owners 
a mail pointing out that while STD0010/RFC0821 doesn't say anything 
about time-outs, RFC2821 does specify that the time-out for DATA 
termination SHOULD be 10 minutes.

> I haven't found a way (if there is) to tell mimedefang not start
> processing a message until after it's been completely received.

Isn't that what mimedefang normally does (unless you use filter_relay 
etc)? When the "." terminating the DATA is received (meaning, the 
message is completely received), mimedefang processes it before 
sendmail replies with the result of the DATA.

If you want the server to not only receive the message but also reply 
to the sender before processing, that's different. If that's not what 
you want, disregard my comments below.

You'd have to change where sendmail calls mimedefang. Now I don't 
know enough about sendmail and milter so I don't know of you can tell 
mimedefang to hook into sendmail just before sendmail queues the 
message but after it has sent it's reply for DATA. A quick look in 
the libmilter docs makes me think that there's only one hook called 
after a complete message is received, and my guess is that that's the 
hook mimedefang allready uses.

You could use two mailservers. One that simply receives, without 
processing and then relays everything to a second server wich uses 
mimedefang (and you should be able to put both thsoe servers in the 
same machine, though you might have to fiddle a bit for that). You'll 
loose the ability to reject SPAM directly during the SMTP session 
with the server sending it to you though.

Regards
/Jonas

-- 
Jonas Eckerman, jonas_lists at frukt.org
http://www.fsdb.org/