First of all, thanks for your input.

I’ve used this approach in several projects. Well, indeed it was almost the same database structure with different metadata configuration, so it was able to hold different entity types in each project (without any changes in the DAO layer code!).

In a near future (hopefully!) I intend to complete this article with a third part to explore this a bit more, but the main idea is that, once relationships are defined on runtime or by means of metadata configuration, it should be possible to do the same with any other kind of data, so many different types of logical entities could be stored in a single physical structure (a set of database tables). Again, performance takes a hit, but there are advantages that, in my opinion, outweight this for many kinds of applications.

Unfortunately these projects are not open source, so as of now I can not publish the code or any specific details, only the general philosophy behind them.

From our point of view, this method can save us an amazing amount of time. Previously, lets say we wanted to add a new type of object: “Monkeys” to “Companies”, it would have taken quite a few hours to program the model, view, controller, actions, struts config, and database. Now all I need to do is define the object in the database, and everything is generated.

This is great for us as we always get little customizations which require adding a new object type, or varying an object, which is incredible easy now.

I would be interested to know if there are any other projects actually using this method? Especially open source ones.

Great article.