-------- Original-Nachricht -------- > Datum: Thu, 23 Aug 2007 14:24:49 +0200 > Von: "Sven Kloe" <[EMAIL PROTECTED]> > An: dspam-users@lists.nuclearelephant.com > Betreff: Re: [dspam-users] wich group to use?
> > > So many groups and no clue which one/ which combination to use :( > > What do you want to do? What is the goal you try to reach by using > groups? > > I want to migrate from Spamassassin to DSPAM in an ISP-environment. > I have many Domains and multiple E-Mailaccounts per Domain. > > There are E-Mailaliases in a LDAP-Tree (I don't know if I can query the > LDAP) > You mean you don't know if you can query the LDAP from within DSPAM? Or do you mean something else? > The first goal is to provide an generic "out of the box" Filter which > could be > improved by user training. > Then I would suggest to use either a global or a merged group. > The next step is to use the LDAP information to maintain email aliases > with > one dictionary and quarantine (maybe shared,managed) > For all users? I would not do that. Maybe one per domain but not a global one. > I think a general (inoculation?) group per Domain improves the > Spam-detection. > Only if the users do train. The inoculation happens when the members of the group do training. > And it should be maintainable with a frontend (not written yet). > Why not using the DSPAM Web-UI? > If I'm right with the above the new qeustions are: > > Is a global or merged group for the "out of the box" Filter better > (performance?)? > When you ask about performance are you talking about speed or the accuracy of the filter? I have not benchmarked but I think that the global group is faster since it will be only consulted if needed while the merged group is merged in real time (and this costs time/performance). I think from the viewpoint of speed any setup is faster the SA. > A user can't be member of shared and inoculation. > According to the documentation: Right. They can't be. > Exists something like a > managed, inoculation group? > I don't think so. But maybe Jonathan knows more? > In general is a shared group or a combination inoculation/ classification > the > better choice? > I personally think a global or merged group is a better choice. I use merged since merged allows me to train periodically the merged group and have influence on the filter even if the user has trained his filter very well. > Is it possible (does it make sense) to nest groups like this: > group1:global:user1,user2,user3 > group2:inoculation:user1,user2 > group3:classification:user1,user2,user3 > > > > Which combinations are succesful in use? > > > > Probably all of them. > > I can't mix shared and inoculation groups, so I can't use every > combination :( > Okay, okay... You can only use the one described in the README. > regards > Sven // Steve -- Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kanns mit allen: http://www.gmx.net/de/go/multimessenger
