It looks like that JPA specification on this subject is largely subject to interpretation, I don't see from where you get such certainty about what it doesn't tell? Maybe I am not in the secrets of the ORM gods like you ?

We need to stop with this point of view of spec interpretation. That is not the way a spec works. If a spec does not say something specifically, then it is open to interpretation. The spec says absolutely nothing about this, as you astutely point out. So we as a team are perfectly fine making our interpretation. You may not like it, but that != "non spec compliant" or any such phrasing. What we do clearly does not violate the spec.

Is it intuitive? Meh, that's open to debate. I can certainly see that point of view, so much so that I planned on adding that ability in 6. But as Vlad points out this has been the behavior in Hibernate well before JPA 1.0 even. It's funny how people hate changes in behavior unless its the one thing they want changed So you have a user base out there relying on the behavior as it is now and always has been. Why on earth do you think we would change that based on this (inaccurate) spec argument?

The Tuple option is very much a viable work around and the entity in the Tuple is in fact the entity instance.

That is exactly where we disagree over the interpretation of the specification.

Exactly - interpretation of the spec. Not "something the spec explicitly says".