> >So the  fix was  very useful,  but it still  seems like  LISTSERV's loop
> >detection for command jobs could  use some improvement, such as limiting
> >the number of command jobs accepted from any one address in a day.
>
> But  this would  have  a number  of undesirable  side  effects. The  real
> problem in your  case is that the commands were  considered valid because
> of the old  I = INFO abbreviation,  which is going away in  1.8c for just
> this reason.  The empty messages,  as I told you  in private this  is the
> first time  in 9 years  that I hear of  a loop involving  empty messages.
> Let's face  it, the loop prevention  will never be perfect.  The incoming
> RFC822  headers  can be  invalid,  for  instance. Should  LISTSERV  start
> maintaining counters  for the  number of  occurrences of  specific RFC822
> errors and  start rejecting  all messages  with a  certain type  of error
> after a certain threshold? :-) I could spend a lifetime and a half adding
> catches for specific problems that slip through the normal detectors, and
> one day there would still be a user figuring a way to screw up anyway.
>
>   Eric
 
What would be the undesirable side effects?  A limit would also catch
some of these turkeys who are subscribing to every list they can
find.
 
People here are depending on LISTSERV now to conduct business and
this problem made LISTSERV nearly useless for two days.  The problem
might have been detected sooner if we had had more people at work
(NIH is mostly shut down due to the U.S. Government funding fiasco),
but the people who were at work were still depending on LISTSERV for
it's normal functions and to distribute information about the
furlough.  It failed miserably.  Anyway, it still took over a day to
recover after the problem was discovered, although to be fair, some
of that is due to the machine being overloaded.  We are trying to
address that, but not this past week.  :-(
 
I don't know exactly what the offending rules were that the Groupwise
user had set up, but I suspect that they were sending at least two
messages in response to each one received.  LISTSERV responded to
each of those, creating an exponential expansion in the number of
messages.  So it was more than just a simple loop.