> Here's a neat little thing:
>
> A guy subscribed to one of our mailing lists via a web interface.
>
> He had an automatic little vacation program set up on his account.
>
> The message telling him that he was subscribed was sent to him.  An
> autoreply was generated.

There are other ways to get such a loop going.  The one we see most often
occurs with regular list subscribers who set vacation messages, usually
with Groupwise systems.  It goes something like this:

(1)  The subscriber gets a message from the list.

(2)  Their mail system sends the vacation message to the list.

(3)  They get another message from the list (perhaps their own vacation
message).

(4)  Their mail system sends the vacation message to the list again.

(5)  LISTSERV detects a duplicate message to the list and sends a warning
message to the user about that.

(6)  The user's mail system sends the vacation message again, but this
time to LISTSERV, not the list.

(7)  LISTSERV attempts to execute the vacation message and sends a
response back to the user.

(8)  The user's mail system sends the vacation message to LISTSERV
again, which attempts to execute it again.

> The autoreply hit listserv and a reply was created telling him that there
> was no such commands was generated.
>
> The 'no such commands' message was sent to him and an autoreply was generated.
>
> The autoreply hit listserv and a reply was created telling him that there
> was no such commands was generated.
>
> The 'no such commands' message was sent to him and an autoreply was generated.

Usually such a loop terminates because LISTSERV will automatically
serve off an address that executes 50 (20 before 1.8c) invalid
commands in a row.  However, occasionally a line of the vacation
message will begin with the name of a valid LISTSERV command.  Only the
first word in a line matters.  Errors encountered later in a command
don't count towards serving an address off automatically.  With
LISTSERV 1.8c that command is usually "THANK", when the user ends the
message with "Thank you".

In earlier versions of LISTSERV, "I" was a valid command (abbreviation
for "INFO").  Many vacation messages started out with "I will be on
vacation until ...". But in LISTSERV 1.8c "I" is no longer a valid
command.

> And so on, until his ISP got overloaded and died, thereby killing his
> autoreply.
>
> It didn't help that this started around 3 am.
>
> The first symptom noticed at our end was that the LISTSERV postmasters
> received 400-odd bounce messages from his account.  The second was that he
> called us up and got kinda irate.
>
> It took us a while to figure out what had happened, so I'm passing this
> along in case any of you find yourselves in a similar situation.
>
> -jwgh

I presume that the ISP's mail server crashing is what led to the bounce
messages.  If that doesn't happen, there are no bounce messages and the
loop goes on until the user gets back from vacation.  We have a program
that scans LISTSERV logs and, among other things, warns when there are
a large number of identical commands from one address.  Such addresses
can be manually served off by a LISTSERV postmaster.

The worst LISTSERV command processor loop we ever had was a couple of
years ago.  An especially clueless Groupwise user had set up a vacation
responder that sent two messages back for each one it received.  So the
number of looping messages increased exponentially.  Unfortunately,
we were running LISTSERV on a underpowered machine then and our machine
died first.