A few comments to some remarks I received about automatic server update:
 
1) It is  true that if I distribute a  bug, it will cause a lot  of servers to
   crash.  However,  you should  remember  that  the  stuff I  distribute  has
   generally been working on my server for at least 4 days. Bugs are therefore
   usually non-lethal and don't kill the server instantly without remiss.
 
2) If the bug was propagated quickly, the fix will propagate even faster :-)
 
3)  I  would  probably  send  updates   more  often  if  they  were  installed
   automatically. At  present I  make sure not  to send more  than 1  update a
   month, except of  course for fixes. Reason: people will  get tired and will
   stop installing them.
 
4) An important advantage is that the server will know (thanks to the protocol
   I'd use) when it has all the  files required to install the new version. If
   you have disabled auto-update, you will then have all the stuff transferred
   to your  account, but  ONLY at  that time. That  is, you  won't be  able to
   install shipment #1  as soon as you  receive it and then  get into problems
   because they  were 2 shipments,  or install  without having got  the UPDxxx
   memo, etc.  LISTSERV will keep  everything in its  reader until it  has the
   complete set of files, and then transfer everything to you.
 
5) A last advantage is that a postmaster can no longer say "I'm not installing
   fixes because I don't have time". :-)
 
 
Now for the drawbacks:
 
1) Technical problems  with receiving updates in the right  order. That can be
   fixed   with   some  sort   of   operand   to   the  update   command,   eg
   PREREQ=1.5i,FIX15I. That's why I don't want to spend time on the program if
   only 5 people are going to enable it: it's something that will take time to
   design and to program.
 
2)  If  I  distribute a  'bug'  that  crashes  all  servers, I  will  be  held
   responsible for  it :-(  :-( Note that  in the case  of servers  running an
   antediluvian version of the code, I spend HOURS writing mail to angry users
   and explaining  why it doesn't  work as documented,  why I can  do nothing,
   etc.
 
3) If there is an error during the CARD LOAD, eg disk full, the server is left
   in an unknown state, where it will perhaps no longer be able to complain if
   it's running under SP3 (under SP4 I'd  EXECLOAD all the execs that might be
   needed to send a 'beep' to the postmasters so no problem).
 
So far,  3 yes, 2  nos (plus 1  that doesn't really count  as it's due  to the
postmaster having local hacks in the code :-) )
 
  Eric