I had a talk with Dirk (sharenet developer) about the use of ACS Objects. What is bad about using ACS Objects as a reference table to the actual tables? We even thought that it should be possible to have acs_object_<package-name> views which could be used for all package related stuff. And all services could either use a Meta View (acs_objects, which stretches over all package views) or implement their own strategy. I never felt very comfortable with the idea of having million objects in one table which will be queried all the time. But if it works, why change.
Another idea that came up, instead of splitting up ACS objects, why not store even more information in it. Minor attributes of objects (e.g. like an urgent_p for bboard postings) could be additional columns to acs_objects, having a NULL value if the object does not have this attribute. According to Dirk indexes will internally drop the NULL columns, therefore not resulting in a speed loss.
Well, maybe I should have joind #openacs and discuss the stuff before posting, but it's late out here and I didnt want to loose these ideas before going to bed.
