Dpsam: Activation, configuration and usage

[fcolpron]

We're currently using Dspam as our unique spam-filtering technique (along with the firewall dropping connections for non-existent email adresses) at some of our client's site, and using Zimbra default spam-filtering configuration at our own location.

We found out that SpamAssassin is somewhat less performant than Dspam, except that SpamAssassin requires very little training. However, our clients are used to Dspam's performance in the range of 99,99% and we know that we won't be able to achieve it with default spam filtering configurations.

So we want to turn Dspam on, but we need to understand how is Dspam implemented to Zimbra, and found very little documentation.

First, according to the Admin Guide, Dspam activation is as easy as:
zmlocalconfig -e amavis_dspam_enabled=true
Is there anything else that needs to be installed or configured?

Are Dspam corpuses for the whole domain, or per user? as we currently have per users corpuses and it's a feature our clients may want to keep.

When moving a message to junk folder, is Dspam trained on top of SpamAssassin?

Ca we give Dspam scores higher values than default ones?

Is there something about Dspam and Zimbra I should know that I'm missing?

TIA.

[uxbod]

I believe the corpuses will be server wide just like the SA Bayes in Zimbra.

[fcolpron]

Thanks Uxbod for the reply, but is there any documentation regarding Dspam implementation in Zimbra, so we can be sure?