On 3/2/12 5:12 PM, Eric Shubert wrote:
On 03/02/2012 04:47 PM, Casey Price wrote:
Not to get too far off track here, but while we are on the topic of
dovecot...has anyone had any experience setting up shared/public
namespaces in dovecot for global spam/ham folders? I'm trying to figure
out the best way to train spamassassin, and the issue at hand is that
the mailboxes are actually stored on a separate QMT machine, and my SA
boxes simply do the scanning/filtering before passing the mail to the
next hop using smtproutes. Just fishing for some ideas.
Let me try this again. ;)
Thanks :-)
I think you're asking for trouble by setting up a mechanism where by
users can train your bayes database, at least in an ISP setting with
multiple domains. What's spam to one person could be ham to another.
With so many users, this could easily get out of hand.
That is a very good point. I guess I wasn't necessarily thinking I'd be
allowing the users to do the training, but that spam would be put
somewhere, and I could periodically run sa-learn against said spam,
although from what you wrote below it sounds like that is even a recipe
for trouble.
To begin with, SA is not intended nor geared for global learning. At
best it can handle relatively small domains. Certainly not
cross-domain learning. When it comes to global learning, DSpam is
better at that. Unfortunately, QMT doesn't do DSpam yet, and don't
expect it soon.
What problem is it that you're trying to address? I think setting up
SA with training in an ISP setting is perhaps a solution that'll end
up being worse than the problem it's intended to solve. I'm not
running an ISP setting though, so I could be wrong.
The problem is that some of my customers are complaining about increased
levels of spam. Some domains get a ton of spam, while others hardly any
at all. I really want to make use of spamdyke as much as I can, as it
really reduces the load on the server and prevents mail from being
accepted and processed that doesn't need to be. Some of the spam is
being tagged by spamassassin, while some is not.
Ultimately, allowing per-domain or even per-user spamassassin configs
would be ideal, but therein lies my problem...spamassassin is running on
a separate server (a server that just knows about each domain it serves,
not which users belong to the domain). So...perhaps my answer is to run
the spamc client on my servers at the last hop, and have them connect to
the spamassassin box.
I'm basically just looking for an effective way to improve my spam
filtering, and allow different configs for different users/domains
without having to restructure my whole system. How are other large hosts
or ISPs doing things?
Any ISP users out here that have an opinion about this?
Casey Price
Smile Global Technical Support
Submit or check trouble tickets http://billing.smileglobal.com
www.smileglobal.com <http://www.smileglobal.com>
Follow us on Twitter <https://twitter.com/#%21/SmileInternet>
Find us on Facebook <https://www.facebook.com/smileglobal>
