On Wed, 27 Oct 1993 08:35:08 CST "George R. Bancks" <[log in to unmask]>
said:
 
>Now if I had  a QUERY subscribed users for my node,  I could perform the
>command only once a week/month.
 
Unfortunately, the overhead  for netwide QUERY is  not significantly less
than for netwide DELETE, once you take into account the fact that netwide
DELETE only needs to be issued once at the end of the semester.
 
In both  cases, all LISTSERVs  have to run a  job, and examine  all their
lists.  In  the case  of  netwide  DELETE, LISTSERV's  caching  mechanism
eliminates the  need to perform any  I/O in most cases,  although in some
cases it  will be necessary.  For a netwide QUERY,  all lists have  to be
read sequentially. This is because hostnames  can have aliases and can be
changed automatically,  and they are thus  not suitable for use  as a key
field.
 
>So we are back to solving problems  after they happen instead of using a
>little foresight and fixing problems before they happen.
 
But I don't see a bounced message  as a "problem". A continuous stream of
bounced messages for the same userid would be a problem, but not a couple
individual ones. Most mail systems let  you forward a copy of all bounces
generated  by your  mailer to  a  local account.  You can  run a  program
periodically that  checks the mail  of this account, detects  cases where
the 'Return-Path:' field is of the type 'owner-xxx', and sends a suitable
SIGNOFF command. It is really not that complicated:
 
Return-Path: <[log in to unmask]>
 
-> Send mail to some.edu with the command 'SIGNOFF HELLO'
 
You will  want to log  that action so that  your program doesn't  send 10
SIGNOFF HELLO commands in a row, but that's about it.
 
  Eric