View Single Post
  #26   Report Post  
Old 09-11-2007, 06:32 PM posted to rec.ponds.moderated
Chris Barnes Chris Barnes is offline
external usenet poster
 
First recorded activity by GardenBanter: Apr 2007
Posts: 81
Default Paying for the moderation software for RPM

Drew Lawson wrote:
Um, just so you know, moderated Usenet groups are entirely based
on an "email type scheme." New posts are diverted to the standard
moderation email address for the group.

These and some other constraints have pretty much
forced us to go with a web-based program for moderation. That
requires a "front end" to the program which can deal with most of the
spam and other stuff that a moderated group is subject to,


I will note, since you are on the threshold of paying for this
again, that your selected software also has a strong tendency toward
false positives with no helpful messages. I have yet to get a post
through to the moderation queue. (Unless this one happens to work.
I keep trying... Nope. Have to hotwire it yet again.)


What you say is true, but just because the messages arrive via email
does not preclude the use of a web-front end for management purposes.


Thinking out loud here....
If I were to design a multi-user moderation system from scratch, I would
probably start with a Linux box running postfix (for receiving email),
Spamassassin (for scanning messages for spam), Squirrelmail with the
Bounce Addon (a webbased email client), and INN or Dnews (usenet server
software). While I'm at it, I would install Mailman too. Note that all
of this is free (except the computer itself to run everything).


(1) Start by setting up a regular user account (called "pondmod") for
the receiving of the incoming email posts for review. Give every
person on the moderation team the password to this account (note 1).


(2) Create 4 additional email folders for the pondmod account:
Spam, Ham, MaybeSpam, & Approved


(3) create a .procmailrc script on the pondmod account which would:
(a) move messages marked as spam (by spamassassin) to the MaybeSpam
folder so a moderator can review them for false positives (note 2).
(b) check the incoming message against a "blacklist" (containing both
email addy and sender's ip addresses). If it matches, move the message
to /dev/null.
(c) check the incoming message against a "whitelist" (containing both
email addy and sender's ip addresses). If it matches, "bounce" the
message to INN for posting to the group.


(4) set a crontab to run every 5 minutes where INN (or Dnews) will look
in the ~pondmod/Maildir/.Approved/cur folder. For each message it finds
there, add the appropriate "Approved:" tag and post to the rpm group
(removing the message from the folder when it's done).


(5) Do the same thing as step 4 for the messages in the Ham folder, but
do NOT remove the message (yet). Instead, run "sa-learn --ham" (part of
spamassassin) on the Ham folder and "sa-learn --spam" on the Spam
folder. Remove the messages in both folders at the completion of this
step. Note that doing this helps improve SpamAssassin's efficiency in
correctly identifying spam.


When one of the moderators wants to review the messages, they simply
logon to the Squirrelmail webpage and look at the messages in the Inbox.
To approve a message, they simply move the message to the Approved
folder. If a message is in the Inbox and is spam (but not marked as
spam), the moderator manually moves it to the Spam folder. The
moderator then needs to look in the "MaybeSpam" folder. Move any
messages which indeed are spam to the Spam folder; move any messages
which are not spam (should be approved) to the Ham folder.

Done this way, the task of moderating a newsgroup is no different than
simply checking one's email on gmail or hotmail.


Note 1: You could give every person on the moderation team their own
account if you wanted to. But if you do this, you'll need to create a
.procmailrc in their home directories pointing at pondmod's Maildir.
You'll also have to play with the permissions of the directory to make
sure they can read/write the files in that directory.

Note 2: The reason for moving messages Spamassassin thinks are spam to
the MaybeSpam folder, rather than just nuking them a
(a) risk of false positives - 1 false positive is worse than 1000 false
negatives. Said another way, you don't want to nuke legit messages.
(b) as already mentioned, running sa-learn helps SpamAssassin do a
better job of identifying spam in the future, resulting in lower false
positive and negatives.


Note 3: probably the hardest task in all of this is getting an existing
usenet server to accept your server's message uploads. You'll need to
find another Usenet server admin willing to accept your connections. On
the other hand, you don't need to worry about getting a download feed
from them at all.

--

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Chris Barnes AOL IM: CNBarnes
Yahoo IM: chrisnbarnes
"Usenet really is all about standing around and hitting the ground
with clubs, on a spot where many years earlier a dead horse lay."