Date: Tue, 8 Aug 2006 21:58:03 -0300 (ADT)
From: "Marc G. Fournier" <scrappy@postgresql.org>
To: Bruce Momjian <bruce@momjian.us>
cc: Josh Berkus <josh@agliodbs.com>, pgsql-www@postgresql.org
Subject: Re: Maia Mailgard down?
In-Reply-To: <200608081435.k78EZ1212146@momjian.us>
Message-ID: <20060808215225.I7267@ganymede.hub.org>
References: <200608081435.k78EZ1212146@momjian.us>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed

On Tue, 8 Aug 2006, Bruce Momjian wrote:

>
> Because all my momjian@postgresql.org email is forwarded to
> bruce@momjian.us, I don't have any way of training the spam filter.  Is
> there a way I can use a global one?

MAIA is *before* your mailbox, so, before the filtering ... the mail 
system essentially goes:

smtpd -> greylist -> maia -> mailbox -> local store
                                             or
                                      -> forward

so, you can still train the system, just login, go to Settings and set 
your appropriate settings so that it Quarantines stuff, and then train 
away ...

The nice thing about the system is that it doesn't just train/report to 
the Bayes database, but it reports it to Razor2, Spamcop and Pyzor as 
well, but, it does it in the background ... the interface just allows you 
to confirm an email as spam, and a cron job then runs to actually do the 
appropriate processing of the email(s) ...

any non-spam that you 'release' is, of course, released instantly back 
into the system ...

there are a few things planned for v1.0.2 that will be nice ... like the 
ability to just auto-discard anything that scores over X (if someone sends 
me an email taht scores over, say, 20, chances are it isn't something I 
want to read or ever really see) ...


  >
> ---------------------------------------------------------------------------
>
> Marc G. Fournier wrote:
>> On Sat, 5 Aug 2006, Josh Berkus wrote:
>>
>>> Folks,
>>>
>>> Is there any way we can return to whatever we were using before Maia
>>> mailgard? Maia is patently NOT working ... it's fundamentally broken in
>>> some way, and isn't blocking any spam.  The interface for Baysian
>>> filtering is also awkward and darned near unusable.  In a couple months
>>> we're going to be trying to do the PostgreSQL 8.2 release which means
>>> that we need the @postgresql.org e-mail addresses working ... and right
>>> now they're all receiving a flood of unfiltered spam, making them
>>> unusable.
>>
>> Actually, all MAIA is is a front-end over what we were using before
>> (amavisd + spamassassin) to allow individual mailboxes to configure
>> themselves, instead of one big site-wide setting ... I'm getting a
>> *trickle* of spam right now, but I've spent the past week or so going
>> through and training the Bayesian database ... in fact, under the old
>> system, those 46 messages that are currently quarantined in your cache
>> would have been delivered, instead of quarantined ... in the
>> @postgresql.org one that I monitor, there are >7k messages currently in
>> spam-quarantine that normally would have been passed through to the system
>> ...
>>
>> Not sure what you are finding awkward about the interface, but if you go
>> in and just turn on all the various checks, but *disable* the
>> quarantining, so that it just label's and passing everything over to you
>> to filter, you will have exactly the same thing in place that was there
>> before we added the interface for per-user settings ...
>>
>> Please note ... at *no time* did we ever reject spam for any mailbox,
>> except those for the mailing lists themselves, which majordomo is setup to
>> do ... so if you are seeing more now then before, its because we are being
>> hit with more now that are passing through spamassassin ...
>>
>> Note that unlike the ~60 messages in the cache for your mailbox right now
>> (12 unconfirmed non-spam, 46 unconfirmed spam), the stuff that I go
>> through is right now sitting at ~10k, and I generally spend about a half
>> hour in the evening going through it ... I generally care about
>> unconfirmed-spam between 5 and 10, ignoring most above that ... and
>> unconfirmed non-spam between 0 and 5 ...
>>
>> ----
>> Marc G. Fournier           Hub.Org Networking Services (http://www.hub.org)
>> Email . scrappy@hub.org                              MSN . scrappy@hub.org
>> Yahoo . yscrappy               Skype: hub.org        ICQ . 7615664
>>
>> ---------------------------(end of broadcast)---------------------------
>> TIP 6: explain analyze is your friend
>
> --
>  Bruce Momjian   bruce@momjian.us
>  EnterpriseDB    http://www.enterprisedb.com
>
>  + If your life is a hard drive, Christ can be your backup. +
>

----
Marc G. Fournier           Hub.Org Networking Services (http://www.hub.org)
Email . scrappy@hub.org                              MSN . scrappy@hub.org
Yahoo . yscrappy               Skype: hub.org        ICQ . 7615664