Multiple repositories in the same OBIEE server instance?

This is a strange one. The service INI file (nqsconfig.ini) for the OBIEE does suggest that the server can handle more than one repository at a time. In fact, if you do add another repository to the configuration file, that file must be a valid repository as well. Otherwise the server won't start.

There is a small thread on the Oracle Forums about this, where someone posts from the docos how this can be achieved using more than one Presentation Server services (an Oracle rep told me that this can be done easily on Linux but harder on Windows). But setting up a new presentation service for each repository doesn't feel right, somehow. Scott Powell mentions in this thread that he has requested (I assume he as real means to deliver such requests) that the repository could simply be referenced in the direct URL. That'd be snazzy, wouldn't it?

In the setup my client has, we have the web-server on another machine than the BI server. The web server is in fact a load-balanced OAS and the BI server (with the presentation service) seems to have no problem with the load-balancing thingamajig (the ops guys were playing with slashing down one web server at a time, to see what would happen to the sessions: The dashboard didn't seem to notice a thing :) ). Control the access is done via LDAP. It took me but 30 minutes or so to find the documentation, create the initialization block described there and log on using my own credentials. So far I have not been able to get the GROUP (or WEBGROUP) access configured. The idea is that the LDAP server will store what groups the logged on user has access to. I created a test group in the BI server and granted that group some read access. As I logged on (and the LDAP values got copied in to the GROUP variable) I still didn't gain access to the desired report. Must be something I did wrong, eh?

I am very keen on using the Answers module now and learning more how to get things done there. The SQL that is generated isn't always what I expect it to be, but most of the time I haven't manage to get what I want. One thing I saw yesterday that made me puzzle a bit was how it used something like analytic min(col1) over (partition by col2) where I wasn't expecting it to, and then a regular max(col3) right next to it (which I wanted). Again, must be something I did wrong somewhere (perhaps in the definition of the fields in the repository?)