All, I was just reviewing https://reviews.apache.org/r/6557/diff/ for the Hibernate -> Custom DAO implementation and it makes me so sad. Its not that you went from Hibernate ORM to custom built ORM (that is already in CloudStack), you went from ORM to raw SQL. From the size of the diff and the amount of lines of code added, its an advertisement in itself in why one would want to use an ORM.
I just joined this mail listing so I probably missed the whole back story, but from what I can gather, you can't use hibernate because of the license and the consensus is to just make it consistent with CloudStack, which already has a custom ORM/DAO implementation. This comment is probably too late in the game, but I really think your missing an opportunity here. For the future of CloudStack I think most everyone would agree that it should move to a standard ORM approach and drop the custom one it has. Using my "Jump to Conclusions Mat" I would say that such a discussion will end up with deciding that Spring+JPA should be the interface/container and the implementation I don't know (I'd vote Eclipse TopLink if the license is compatible). So the AWS component is a perfect place to test out such framework do to its isolated nature. Having a lot of experience with hibernate/jpa and CloudStack I have a good idea in my mind just how difficult it would be to convert all of the CloudStack code to a JPA compatible solution. If you were to do the conversion, for practical reasons, you are going to have to do a phased migration. This means that you'd have to support old framework and new framework concurrently for sometime. Coming up with an approach to reconcile the two frameworks will take some time and effort. Since the EC2 component is a pretty well isolated framework, it would be easy to implement Spring/JPA and test the validity of what would be the end state before you take on the complex and long task of converting all of CloudStack. I'd even be willing to work on this. It just seems silly to me to go from ORM to SQL and then back to ORM. But I realize, as I said before, this comment is somewhat late in the game. Darren
