Barry, what I meant was that a function like get_attribute works great for selecting a single object attribute. Assume a message object with attributes title, body. If you wanted a summary of messages, a list of titles, it would be more efficient to know where the object attribute is stored...so you can write an efficient query to return a list, instead of writing a loop of get_attribute...Versioned storage is the most complicated to work with. Having non-versioned storage would be easier, generic storage, easier still, although less efficient. With generic storage you wouldn't need to know anything about how storage was handled since object metadata would be used to return results. Versioning could also be folded into generic storage. Issues with generic storage are inefficiency and the difficulty of data integrety, since data typing and references would be difficult to establish.
