Thanks, Thomas and Rolf, for your advice, and apologies for troubling you. I have fixed the problem.

Although my domains in mercury.ini didn't have the errors suggested, and the structure of the entries for my new domain replicated the existing (and working) ones, there was a surplus line in them - dating from the time when I first set up Mercury and knew even less than I know now.

I have learned two things which I hope may be of use to others

- if your domains are not set up correctly you can get weird things happening, such as test@domain.com being a valid address and test1@domain.com being invalid, even with a domain mailbox. Such effects may be completely random, and not symmetrical between accounts, even if the accounts are set up the same way.

- don't munge (edit) log postings to try to avoid publishing domain names or addresses (not even a little bit)

 Chris

I've just set up a new Mailbox on my Mercury server, with associated domain (and domain mailbox). To test, I have sent mail from another user on the server, which is received normally. Mail from the new user to any other user, however, is giving me an error. Mail to externals servers works fine.

The error is "Job has invalid or illegal from address". The 'from' address is the one which receives mail without problem, and is a normal address of the form user@domain.co.uk  I've checked it and rechecked it, and retyped it from 2 separate clients and it looks normal.

The error occurs for several variations of user on this domain - I have not found any which work - and only occurs from this new mailbox/domain.

The error appears in the core process window. If validation of from address is suppressed, delivery then occurs normally. If it's not suppressed, there's a stream of messages in the core process window, failure to deliver and two errors to the postmaster. The first error says that the message has an illegal 'from' address, but confusingly then states the 'To' address. The second error results from trying to deliver an error to the sender, and says the address is not known - despite the fact that normal mail to the address is delivered.

The error occurs whether I send the mail directly using MercuryE or directly through my ISP's relay. I use MercuryI to deliver the mail to the client (Thunderbird)

The only difference I can think of between the new domain and the existing ones is that I haven't yet had the MX records transferred for the new domain, so incoming mail goes to the hosting company, where it is forwarded to alias on an existing domain and dropped into the correct mailbox via a filter. But the problem is when I'm sending from this domain, not receiving.

- Restart Mercury to make sure all changes have been updated in the program.

- If you still have the problem, check the Local domains section of core configuration carefully.

/Rolf 

The error appears in the core process window. If validation of from

address is suppressed, delivery then occurs normally. If it's not

suppressed, there's a stream of messages in the core process window,

failure to deliver and two errors to the postmaster. The first error

says that the message has an illegal 'from' address, but confusingly

then states the 'To' address. The second error results from trying to

deliver an error to the sender, and says the address is not known -

despite the fact that normal mail to the address is delivered.

I'd really like to see a copy of what is sent to either MercuryS or the mail spool directory.  Depending on how the message is sent either turn on session logging in MercuryS or pause core and show us a copy of the 101 file.

Rolf, thank you; I have already tried both those things.

Thomas, thank you. I haven't been able to find the 101 file, but here is the log from MercuryS (not munged except user addresses)

20:46:51.640: Connection from 212.159.7.35, Wed Jul 15 20:46:51 2009<lf>
20:46:51.765: << 220 cjbolton.plus.com ESMTP server ready.<cr><lf>
20:46:51.843: >> EHLO relay.ptn-ipout01.plus.net<cr><lf>
20:46:51.875: << 250-cjbolton.plus.com Hello relay.ptn-ipout01.plus.net; ESMTPs are:<cr><lf>250-TIME<cr><lf>
20:46:51.890: << 250-SIZE 0<cr><lf>
20:46:51.906: << 250-AUTH CRAM-MD5 LOGIN<cr><lf>
20:46:51.921: << 250-AUTH=LOGIN<cr><lf>
20:46:51.921: << 250 HELP<cr><lf>
20:46:51.078: >> MAIL FROM:<xxx@andrewbolton.co.uk> SIZE=803<cr><lf>
20:46:51.296: << 250 Sender and size (803) OK - send RCPTs.<cr><lf>
20:46:51.375: >> RCPT TO:<xxx@cbolton.co.uk><cr><lf>
20:46:51.390: << 250 Recipient OK - send RCPT or DATA.<cr><lf>
20:46:51.437: >> DATA<cr><lf>
20:46:52.578: << 354 OK, send data, end with CRLF.CRLF<cr><lf>
20:46:52.640: >> X-IronPort-Anti-Spam-Filtered: true<cr><lf>
20:46:52.656: >> X-IronPort-Anti-Spam-Result: Aq0EAJvOXUrUnw6T/2dsb2JhbACSM8B0hAkF<cr><lf>
20:46:52.656: >> Received: from ptb-relay03.plus.net ([212.159.14.147])<cr><lf>
20:46:52.671: >>   by relay.ptn-ipout01.plus.net with ESMTP; 15 Jul 2009 20:46:05 +0100<cr><lf>
20:46:52.671: >> Received: from cjbolton.plus.com ([212.159.73.44] helo=[127.0.0.1])<cr><lf>
20:46:52.687: >>      by ptb-relay03.plus.net with esmtp (Exim) id 1MRAQX-0005pI-8k<cr><lf>
20:46:52.703: >>     for xxx@cbolton.co.uk; Wed, 15 Jul 2009 20:46:05 +0100<cr><lf>
20:46:52.703: >> Message-ID: <4A5E3229.2070409@andrewbolton.co.uk><cr><lf>
20:46:52.718: >> Date: Wed, 15 Jul 2009 20:46:49 +0100<cr><lf>
20:46:52.734: >> From: Andrew <xxx@andrewbolton.co.uk><cr><lf>
20:46:52.734: >> User-Agent: Thunderbird 2.0.0.22 (Windows/20090605)<cr><lf>
20:46:52.750: >> MIME-Version: 1.0<cr><lf>
20:46:52.765: >> To: xxx@cbolton.co.uk<cr><lf>
20:46:52.765: >> Subject: Test<cr><lf>
20:46:52.781: >> Content-Type: text/plain; charset=ISO-8859-1; format=flowed<cr><lf>
20:46:52.796: >> Content-Transfer-Encoding: 7bit<cr><lf>
20:46:52.796: >> X-Antivirus: avast! (VPS 090715-0, 15/07/2009), Outbound message<cr><lf>
20:46:52.812: >> X-Antivirus-Status: Clean<cr><lf>
20:46:52.828: >> X-Plusnet-Relay: 654c4c314561163df27deb65e4f2c874<cr><lf>
20:46:52.843: >> <cr><lf>
20:46:52.843: >> <cr><lf>
20:46:52.859: >> .<cr><lf>
20:46:52.875: << 250 Data received OK.<cr><lf>
20:46:57.921: >> QUIT<cr><lf>
20:46:57.937: << 221 cjbolton.plus.com Service closing channel.<cr><lf>
20:46:57.953: --- Connection closed normally at Wed Jul 15 20:46:57 2009. ---

Despite 20:46:51.296: << 250 Sender and size (803) OK - send RCPTs.<cr><lf>  the postmaster then received

Received: from spooler by cjbolton.plus.com (Mercury/32 v4.72); 15 Jul 2009 20:47:13 +0100
X-Envelope-To: Electronic Postmaster <xxx@cjbolton.plus.com>
To: Electronic Postmaster <xxx@cjbolton.plus.com>
From: Electronic Postmaster <postmaster @cjbolton.plus.com>
Date: Wed, 15 Jul 2009 20:47:02 +0100
Subject: Postmaster Notify: Delivery Failure.
MIME-Version: 1.0
Content-type: multipart/mixed; boundary=22036.79390984

This message is in MIME format. If you are seeing this text,
then your mailer does not understand this format properly.

--22036.79390984
Content-type: Text/Plain; charset=US-ASCII
Content-Disposition: Inline
Content-Description: Reason for delivery failure.

The attached message has failed delivery and has been referred
to you as postmaster. The following error report or reports
were given to explain the problem:


   --- Problems not related to specific addresses in the message:
   Job has invalid or illegal from address.
   *** <xxx@cbolton.co.uk>

--22036.79390984
Content-type: Message/RFC822

Return-path: <xxx@andrewbolton.co.uk> 

[message text followed] 

I have not worked out how to log the core process.

I have created two more accounts as tests - one behaves like this one, the other is fine. I'm still trying out ideas to see why.

Thomas, thank you. I haven't been able to find the 101 file, but here

is the log from MercuryS (not munged except user addresses)

Please, please do not munge a session log except for a password.  In this case it was the most critical part of the session log as well.

[quote user="Thomas R. Stephenson"]

Please, please do not munge a session log except for a password.  In this case it was the most critical part of the session log as well.

[/quote]

Sorry - I took care not to munge any of the domains in the addresses so I though they were still distinguishable. I only changed the bit before the @. If it's critical to know the full email address of the user perhaps I can create a new address that replicates the error, post that and bin it when the discussion is over.

I realise there's no point munging domains as they are public anyway, but surely posting a complete, valid email address is asking to be spammed - particularly when it's a new address and one of the reasons for using it is that the old domain has become so flooded with spam as to be unusable?

In a case like this the Domains section of Mercury.ini should contain a line like this:

DM=domainmailboxname: cbolton.co.uk

But it must not contain a line like this: 

servername: cbolton.co.uk

/Rolf 

I realise there's no point munging domains as they are public anyway,

but surely posting a complete, valid email address is asking to be

spammed - particularly when it's a new address and one of the reasons

for using it is that the old domain has become so flooded with spam as

to be unusable?

Real data is necessary.  There may be a high bit in the users address that is not apparent including a trailing or leading non-printable character.  If an e-mail address is used at all it will be spammed, period.  If it is used in a completely closed system and/or the e-mail address never gets to the internet, posting an e-mail address to this list makes not iota of difference as to whether or not the spammers get the address.

My e-mail (stephens@bayarea.net) address at my ISP has never been used to send anything except to the ISP when I started, no mail ever sent to the internet.  A dictionary attack on my ISP's system found that address in less than a week and now it's being buried in spam.  I download and delete it but I only do that so I can test Spamhalter and POPFileD.  ;-)

More testing has established that there are some usernames at the domain concerned which are treated as valid, and some as invalid. Length of name doesn't seem to be significant. Tried a different client machine to eliminate a corrupt client and it made no difference. Thinking of trying a different client (I normally use Thunderbird) to see if that makes any difference - maybe Thunderbird is introducing a spurious character?

The session logs from MercuryS are the same whether or not the address is flagged as invalid - MercuryS always logs that the address is OK, so it must be the core which is rejecting it.

Rolf, thanks, my domains look OK on the basis of your advice.

Once I have a better picture of what is going on I will post logs (verbatim!) but no point while I'm still gathering evidence.

If cbolton.co.uk is set up as a domain mailbox and only that in Mercury.ini domains section then the part of the address before the @ won't be evaluated by Mercury. The only exception would be if basic syntax rules for email addresses are broken, but in that case I doubt that the message would be accepted by MercuryS in the first place. Could you quote just one address that gets rejected?

/Rolf