Hi All, My personal experience with the current server.xml has been while producing my own distribution of ApacheDS with some of my own custom interceptors and partitions. I found it very powerful and being able to play around with the low level structure of ApacheDS without making changes to the source was very useful. For distribution however, I am making my own configuration file format which will only offer our users the ability to change basic options such as administrator password and listening port number. The distribution is for a very specific purpose and the users should not be messing around with the internals. The server.xml will still be there but it will be hidden away as just another data file rather than something the user is expected to edit.
I wonder if ApacheDS itself shouldn't be doing the same thing? If a systems administrator just wants an LDAP server they shouldn't have to be editing a large XML file with lots of low-level implementation details such as the ordering of interceptors. Perhaps we should have a simple configuration file for administrators that just want to get a server up and running, and a separate one for developers that gives the full power of the current server.xml. To me the xbeans make the file a little easier to read but I think the biggest problem is the structure and sheer amount of configuration that we make available. As for moving the configuration into the DIT - I think that would be great as it would instantly give several nice features: global configuration for all replicas, a way to manage the server remotely, and possibly "live" configuration (no restart required). I do, however, agree with others that this would be losing the ability to hand edit if things go wrong and makes it a little harder to backup. Perhaps we could have a configuration partition which reads from / writes to the XML configuration file(s)? I think the XML configuration files could map almost directly into LDAP entries. I think this would give us the best of both worlds. Cheers, Martin
