OK, let's see what other folks think about a named_object subtype ...
I do, of course, worry about extra joins. However one of the things I want to work on (or get someone else to work on) is the caching of the results of content-retrieval queries in a generalized, consistent manner. Maintaining cache consistency system-wide is a difficult problem but I think that it's solvable in a general way for subsystems like the CR. And that will do a *lot* more to improve performance than removal of a simple join. You can put 2GB ECC Registered DDR into a server for just over $400 - in this day and age we are crazy-stupid for not providing smart content caching where we can do so safely and consistently!
So I'd like to see us work towards a clean datamodel that, of course, is still efficient but also to keep in mind that they BEST way to make db queries fast is to NOT DO THEM. Cache!
