This message is a little long, and if you want to cut to the chase you can
skip to my concluding suggestion at the end ...
 
I've been trying to figure out how LISTSERV could avoid redistributing
commands and error messages sent to the list address (mostly because of
the recent discussion here on that topic, into which I mistakenly injected
a foolish comment I'd meant to send privately ... sigh).  The conclusion
I've come to is that detecting such messages is not at all easy. <surprise>
 
But, I got a little farther than that. :) Thinking back to biostatistics,
it seems to me that deciding whether or not to reject a message is like
blood tests, at least in that results ("reject/post" or "sick/not sick")
can't *always* match up with The Truth.  There *will* be times a bogus
message gets past LISTSERV, same as blood tests sometimes give an
all-clear when the person tested truly has disease.  (A "false negative,"
in either case.) Conversely, there will be times LISTSERV bounces a
message that's legitimate, same as people are occasionally told they have
AIDS when it's not true. (A "false positive.")
 
*Every* test, including LISTSERV's test for bogus messages, has to be
optimized for an acceptable mix of false negatives and false positives ...
decreasing one error rate, barring a jump in technology, means you have to
increase the other.  My point here is that maybe Eric can't go much
farther on reducing the false negatives (that is, the bad messages that
get through) before the false positives start going up in response
(legitimate messages start getting bounced).  And ... maybe that's OK, at
least for some lists.  But for others, it's probably unacceptable.
 
So, here's my SUGGESTION:  what if LISTSERV had several levels of checking
for problem messages, from "trusting" to "skeptic,"  choosable as usual in
the list header.  "Skeptic"  would bounce a lot of messages that didn't
bounce before, and this would be at a cost to subscribers in convenience.
To reduce this "cost," LISTSERV would tell a user when replying to such a
message, rather than what it does:
 
> ... if it was indeed a command you were attempting to issue, please send
> it again to LISTSERV@[blah] for execution. Otherwise, please accept our
> apologies and try to rewrite the message with a slightly different wording
> - possible solutions include changing the first word of the message,
> quoting it, inserting a line of dashes at the beginning, etc.
 
... something like "Otherwise, please accept our apologies and re-send the
message with this line at the beginning:  'Dear LISTSERV: This message is
legitimate ... please don't return it.'"  (you get the point: put in
something that indicates to LISTSERV "I've considered the possibility you
might think this is a command, and I'm telling you it's not.")
 
I think this would be reasonable -- it would allow the use of a *very*
restrictive filter on lists where that was appropriate, yet it still would
let people do postings that "looked like commands" after they'd made sure
they were doing what they meant to.
 
This is one of my brainwaves, but it's late at night and I'm willing for
the idea to be shot down :) ... let me know what you think.
 
Norm
--
Norm Aleks - [log in to unmask] - UMASS Medical Center, Worcester, Mass.
  "For a generation, the most important gay march has been a long line of men
   and women coming out, one at a time."  --Ellen Goodman