[sclug] duplicate mail

[Graham]

Heres my timeout theory:

+ LIST MACHINE CONNECT to your primary MX
- 220 MX-ONE.dom ESMTP
+ EHLO [1.2.3.4]
- 250-MXONE.dom
- 250-ETC
- 250 ...
+ MAIL FROM: <list at server>
- 250 sender ... ok
+ RCPT TO: <you at yours>
- 250 recipient <...> ok
+ DATA
- 354 go ahead
+ Subject: ...
+ From: <sclug-request at sclug.org.uk>
+ To: sclug at sclug.org.uk
+ ... LOTS ...
+ Then the DATA terminating DOT
+ .

<-- here the sender is waiting
<-- for the sever to say "250 ok: Message 1234 accepted"
<-- here the server is overloaded by SPAM from absent taiwanese friends
<-- so it takes more than 2 minutes to respond to the DOT
<-- That client was tested without SO_KEEPALIVE ((guess))
<-- Or some other network drop (such as NAT firewall)
<-- So when the socket is dropped the client didnt get the 250 OK
<-- and it then tries MX-TWO, immediately or later
<-- MX-ONE has succeeded, and sends 250 OK but too late
<-- MX-ONE cannot then undo its own send, or queue-action
<-- MX-TWO succeeds faster (same time) or BOTH delayed downstream
<-- as would happen in an overloaded email system
<-- so they both arrive at the same time, or apart

- 250 ok: Message 1234 accepted

> Message: 4
> Date: Wed, 27 Jun 2007 10:13:05 +0100 (BST)
> From: "Darren Davison" <darren at davisononline.org>
> Subject: [sclug] duplicate mail
> To: "SCLUG" <sclug at sclug.org.uk>
> Message-ID:
> 	<60729.84.92.95.61.1182935585.squirrel at davisononline.org.secondary-mx.co.uk>
> 	
> Content-Type: text/plain;charset=iso-8859-1
> 
> does anyone know why I might be receiving duplicate copies of most (but not
> all) sclug list mail?  Sometimes the duplicate arrives some minutes later than
> the original, often they arrive simultaneously.
> 
> Anyone else seeing similar?

TinCan TalkTalk use a bit of string between two ends.
It often has complete outages, and one-week overflows!
But it no-longer has duplicates. Evolution does though ...

--
Graham