From jkl@ccl.net Wed Jul 15 23:07:16 1992
Date: Wed, 15 Jul 1992 23:07:10 -0400
From: jkl@ccl.net
To: tchemistry@ccl.net
Subject: More problems on the list
Status: R

Dear Computational Chemists,
Two disasters in a week is a lot, especially, when the list was rather stable
for over a year. To be frank, I am human, I even had thoughts about ending this
game. 
What happened? As you know, there are precise protocols which describe how
machines should talk to each other. There are many fields in your header
message which you get from the list. Most important being:
= From: somebody@something --- tells who sent the message
= To: somebody@something --- where the message was sent (in our case, the
                           chemistry@ccl.net)
= Errors-To: somebody@something --- where the error messages should be sent
                                  if machine has problems delivering mail
= Precedence: bulk --- that this is a massive mailing and errors should never
                     bounce to the From: line. Mainly for "vacation"
                     programs which should not send automatic answers.
= Sender: somebody@something --- who actually was sending the message
                     (in our case, chemstry-request@ccl.net)
= Subject: the synopsis of the message for fast scanning which is
        set by the originator of the message.
and some other fields which tell what was the history of the message:
= Received:, X-Envelope-To:, Message-Id:, Date:, Via:, X-Vms-To:, etc.
(Note how I protected reserved fields not to confuse weak minded machines,
I advise you to do the same always, and NEVER put a period on a line by
itself. Some smarter programs can beat this, but some cannot.

So, a correctly configured machine should send errors (i.e., bounced
messages) to the address in the Errors-To: field if it exists, then to
a Sender: if it exists, and finaly to a From: address if other two are
absent. This is a saint thing, on which the soundness of electronic mail
is based.

Some machine in Belgium got crazy (or rather some system administrator did not
go to school) and after silently collecting error messages for a few days,
it spat them back, guess where? To a To: address, i.e., chemistry@ccl.net.
using an old-fashioned Resend-To: field. Why it did so?, Is there any logic
in doing so? You guess the answer... Why send a message again to the place
it was sent before? Beats me...
What this creates? That all of you got a bounced message. But then
the infinite loop is created, and it bounces bounced messages to be
bounced again. I had to stop the list !!!. But things on the network
happen fast, (once it is sent, it is sent...) and you got a lot of bounced
mail (If it helps you, I got ten times more).  Moreover, the wrongly addressed
bounced messages (the header was internaly inconsistent) confused some other
machines on the network, which started to send the junk to chemistry@ccl.net
(but it was then stopped, i.e. redirected to me, so you did not see anything).
We are working now on improving the mailing scripts which will try to
catch inconsistences in the messages sent to chemistry@ccl.net. 
But believe me, it is not easy: if something is unstandard, it can be
anything. And if something is sent to chemistry@ccl.net it is sent to
chemistry@ccl.net...
Please, do not advise me to censor the list. Yes, it would prevent this
>from happening... But I will not do it PERIOD. I have important reasons
not to do it PERIOD.
In a while I will be sending "good" messages which accumulated when the list
was stopped. Please give me a day before sending something to the list,
so I can catch up. I will tell you when. Thanks to Lord, the backlog of
messages is not that big.
If you want to contact me, i.e., you have a REQUEST, send message to
jkl@ccl.net or to CHEMISTRY-REQUEST@ccl.net

If you have to send your message to the list NOW, please send the messages
to CHEMISTRY@ccl.net, as usual, and do not pay attention to the 
sender address on your messages for a few days.

Jan Labanowski
Ohio Supercomputer Center
jkl@ccl.net