On 03/13/2012 11:05 AM, Les Mikesell wrote:
On Tue, Mar 13, 2012 at 12:00 PM, Nataraj incoming-centos@rjl.com wrote:
Ok, so it wouldn't work to just use the oldest received, but a smarter inspection could check to see weather it actually passed through a server owned by the claimed domain. The reality is that what is need is to input this into a scoring system weighted with other spam evaluation mechanisms, something like spamassassin. The downside of spamassasin is that it is costly to run and must be run after the message is accepted by the smtp server.
MimeDefang runs it as part of its own milter process so you don't start a new perl for each message (and multiplexed so there is not necessarily an instance for every mailer process), and at a point where you can do an smtp reject based on the content.
There already exist so many different spam control methods, many of them can run at the smtp level and reject mail prior to accepting. I get pretty decent rejection from greylisting.
MimeDefang doesn't do greylisting although the hooks are there if you wanted to do it yourself. But, you can run multiple milters and milter-greylist works OK.
I would look at the milter that Les mentioned. I haven't had a a chance yet.
I set up the package from rpmforge a long time ago along with their clamav. I see they are both in epel now - not sure what you would have to do to make them work with postfix, though. I've seen them mentioned together on the mimedefang mail list, so someone must be doing it.
I run clamav and the greylisting implementation and the spamassasin interface that's included with vpostmaster. It's not perfect, but it's all quite nicely coded in python and easy to tweak. The GUI allows changing spamassasin, greylisting, clamav parameters on a per user basis. Users can also login to the gui and manage their own spam control. The database also has designed in, the ability to store per user configuration for other modules that you might add yourself. I will look at MimeDefang.
Nataraj