I know what you mean about flip-flopping. I was initially against it when I first heard that we might use pg's inheritance. This was mostly because I had not been hearing good things about it on the pg hackers list. Then I played around with it some, and I thought well it is kind of slick and it does seem to work pretty well. Then after starting to look at the acs 4.x object model, I realized that it was a good match, and I was more in favor of using it, but I was still concerned about the extra porting effort. Now finally that I've looked at a lot of code in the core, I think maybe that it is overkill and not worth the effort. I think I need a more compelling arguement to be convinced that it is worth the effort. I'm going to spend some time looking at some of the new acs 4.x packages and see how much use they make of the acs_objects system, and if I have time I'll set up some benchmarks between the two methods and see if I can better quantify the performance gain.
