On Tue, 3 Aug 1993 16:33:24 -0400 Una Smith <[log in to unmask]> said:
>(1) Automatic error messages should include the address of a real
>person.
The list owner can be reached at [log in to unmask]
>(2) Bounces from a mailing list should go to the list owner.
Well, if I send a message to a list and it is rejected, I want to know
about it. Otherwise I will have to set the REPRO option and waste time
throwing away copies of my own messages in order to know if my postings
make it, and even so I may fail to notice that the message was rejected
since I will be used to throwing away copies of my own messages
routinely, without thinking about it.
>I do not want to disable the checking feature, because this is the main
>way that "loops" are prevented from propagating through the mailing
>list.
Loops are not prevented by the duplicate checker, because in a loop each
message is different (they get longer and longer). The duplicate checker
prevents the list from being flooded by multiple copies of the same
message if a mailer goes wild and decides to send 200 copies of
everything. It also prevents loops between mailing lists and
misconfigured news gateways. A student here recently decided to gateway a
highly sensitive local list we have for higher management types (the
contents aren't confidential or anything, but these people are quite a
bit sceptical about computers). He screwed up, and of course he hadn't
even considered informing me, so the news gateway started sending a copy
of everything it got back to the list while I was fast asleep. Without
the duplicate code, the administrative types would have been flooded with
copies of the same messages, and that would have been the last time they
tried to use a computer for electronic conferencing (it was the first
time this was being attempted, and since it was highly successful they
are going to keep doing that in the future).
>Note, however, that because the loopcheck algorithm sends duplicate
>articles back to the sender, it doesn't prevent loops from developing
>between LISTSERV and the offending account!
The offending account is usually a human being and, when it is not, there
are other mechanisms that stop the loop after a certain amount of
iterations.
>Ideally, I think duplicate submissions should be sent to the list owner,
>who could then deal directly and in a context-sensitive way with the
>original sender.
Or could just hit the discard key because he maintains 50 lists, and the
poster would never know the message was rejected. I cannot take the
responsibility of deciding that all owners will dutifully inform the
poster when this happens. For every conscientious owner, there are two
who simply haven't got the time to take care of such things.
>If duplicate submissions were redirected to the SCIFAQ-L owner (me), I
>would probably not re-send them to the list, but I would want the
>option, as list owner, of doing so without having to modify the text or
>disable the loopcheck feature.
I have no problem with the concept, but how do you intend to tell
LISTSERV to accept the message as it is, without any alteration? Chances
are this will be a highly complicated and unusual procedure that almost
nobody is going to use. Anything based on adding a header field is
unavailable to hundreds of thousands of PC/Mac/VMSmail users. Anything
based on adding some command in the mail body is going to open a can of
worms.
>Also, I wish the loopcheck feature worked within a specified timeframe.
That is a good idea but I'm not sure it can be implemented in 1.8a. The
code that processes postings to mailing lists will be converted to PASCAL
with version 1.8b and then it will be easy to do - and I won't have to
write new code that will be trashed in the next release and rewritten
from scratch in another language.
Eric
|