Great Circle Associates Majordomo-Users
(January 1996)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: Desireable changes to maillist programs
From: "Alan Millar" <amillar @ bolis . sf-bay . org>
Organization: The Bolis Group
Date: Mon, 15 Jan 1996 23:53:11 -800
To: majordomo-users @ GreatCircle . COM
Cc: majordomo-workers @ greatcircle . com
Comments: Authenticated sender is <amillar@hock.bolis.sf-bay.org>

> command. Only check the first line. If it is a properly formatted 
> command then treat the whole message as nothing but commands. If it 
> isn't a properly formatted command then treat the message as text. 

You seem to be saying is that *some* of the following should be 
processed by majordomo and the *rest* forwarded to the list:

1   subscribe listname me@mysite
2   subscribe listname listname me@mysite
3   please remove me from the list
4   subscribe please
5   signoff listname
6   I unsubscribed from that other list and subscribed here because...
7   get listname index.txt
8   get a life
9   please add me to your list.  Thank you.

Now, which of these are you going to have Majordomo process?  What do
you propose to do with the rest of them?  Does majordomo check for
entirely correct lines?  1 and 7 get processed, but 2, 4 and 8 go to
the list.  Oops, we don't want 2 & 4 to go to the list, so we just
check for commands and return errors if the arguments are wrong.
Ooops, we processed 8 as a command: false positive.  And what about
3, 6 and 9? 

I'm certainly not going to run my lists in a manner that *intends* to
let the unrecognized entries above be purposely sent to the list. 
My list members would be pretty incensed by that.  I wouldn't
subscribe to such a list either, without complaining to the list
owner about their lame software.

> IMO this would get around a lot of the problems I've had with 
> administrivia getting too many false positives, as well as reducing 
> administrative overhead.

Too many false positives?  Then take out some of the checks in
resend.pl.  What is the right number of checks?  Too few and more
garbage goes to the list.  You will eliminate most of your false
positives by just removing the word "help" from the checks.  (Chan:
I vote to remove "help" from the standard install.  Just my two
cents :-)

Telling people to send their commands to the list address just
*won't* help your problem.  If they can't get the address right,
they won't get the syntax right either.  Anyone who can [be educated
to] properly format a command can also [be educated to] send it to
the right address.  

Now, if you really believe that users should be rewarded for only
getting one out of two right between addresses and syntax, you're
picking the wrong one.  When they still can't get it right, you
penalize the entire list by forwarding their malformed requests on to
the list.  And it seems that not everyone who can [be educated to]
send it to the right address can as easily [be educated to] properly
format a command.

A better approach is to be more forgiving on the syntax as long as
they get the address right.  I am working on plans to create a
free-form request parser that will sit on the -request address and
look for keywords like resend's administrivia checks.  If it decides
it is a subscribe or unsubscribe request, it will forward a properly
formed request to majordomo, if not then it will send a helpful reply
to the sender.  The MAJOR difference here is that if they *still* put
in garbage the program can't understand (which *will* happen), the
rest of the list doesn't see it!

Has anyone else thought about or worked on a free-form request 
handler, even (gasp!) non-majordomo?  I'd be interested in discussing 
it.

ObMajordomo-workers:  Much of my work went away when I put in an
autoresponder in resend.pl.  When the administrivia checks are
triggered, it automatically sends a message back to the user showing
them the line that triggered it and telling them that how to
subscribe or unsubscribe in case that's what they were asking for. I
can't believe nobody else is doing this! I've been using it for
about nine months, and almost everyone gets their subscription
request right the second time after they get the administrivia reply
from resend.  Note that the offending message is still bounced as
usual; if it is a false positive (infrequent if you take out "help")
I'll simply approve it, otherwise just ignore it.

Chan: is an auto-reply for the administrivia checks in the latest 
version?  Do you want my patch?

The administrivia auto-reply, along with a free-form parser on
-request, will allow users to send crap anywhere and get subscribed
while still educating them on the correct procedures.  This will
benefit *other* list managers as well by reinforcing consistent
standards, and it will benefit all list members who don't want to see
administrivia on the list.

- Alan
----
Alan Millar              amillar@bolis.SF-Bay.org
System Administrator     http://www.bolis.com
I am the great Oz!  Pay no attention to that man behind the curtain!


Indexed By Date Previous: Cross-posting to news group
From: tomkon@borris.khoros.unm.edu (Tom Konerth)
Next: Archive aliases
From: Paul Key <paul@reginet.bham.ac.uk>
Indexed By Thread Previous: Re: Desireable changes to maillist programs
From: "David S. Miller" <davem@caip.rutgers.edu>
Next: Freelance help installing majordomo
From: mark@metrobeat01.metrobeat.com (Mark Davies)

Google
 
Search Internet Search www.greatcircle.com