Quote:
|
Originally Posted by marcmac What would you like to see wrt to spamassassin? DSPAM is on our list (esp. since there's a howto in the forums  )
Spell checking in its own server was done on purpose - we're planning to use the apache instance as an integration point for several services, not just aspell - and, with it in a separate server, you can scale these services apart from the mailbox server, which wouldn't be possible if it was wrapped in the tomcat instance. |
There are still a couple of things that need to be ironed out with DSPAM, training is a problem (Is there any way in Zimbra to associate a account with a command in UNIX?). In addition the DSPAM howto only works for a site wide database. Per user database doesn't work correctly yet (I think it has to do with the Trusted Users, but I am still not seeing independend databases per user).
My ideas were as follows:
- Develop a standard webservice backend API that contains generic Spam Functionality (Run a specific message thru the spam scoring proccess, rescore a message as SPAM or HAM, modify certain generic paramaters - cutoffs, etc, purge the SPAM database if it is no good).
- That backend should be flexable enough to work with different SPAM products, like spamassassin and DSPAM in the opensource world. DSPAM can store user preferences in either a mysql database, or a LDAP instance. The backend should be able to manage thoose preferences and present a single generic API to clients.
- The administration server needs a simple page to control site wide thresholds, generic training and the ability to override user preferences.
- Users need a away to specify their own threasholds, do their own training, and correcting of the spam proccess.
Anyways, I was planning on working on that as my first contribution to Zimbra.