EJB 3.0 / Data control binding / ADF Faces

Question

We have a great and very useful tool to generate rich client web application. Generate web component from a data structure is easy but manage the full life cycle is a little bit more complex. This complexity is dramatically increased when tool/platform are offering many models to solve the problem.
In JDev we have three major choices:
1) adf business components
2) top link pojo
3) ejb3.0
From the client side and in my point of view (at my current understanding) the first and second choice are the easiest way to manage the persistence because they are what i say "bidirectional", that mean i will use the same data structure to persist and synchronize the database. I may only calll a method from the object to put my changes from the midle-tier into the database.
EJB 3.0 seem to be more cormplex due essentially to the presence of the session bean that acts like conversational api. I like this architecture because it is the most flexible but from my client i have to manage these api's instead of one data structure. This mean for me to catch/detect every updates, new insert or deletes in my business object and to write the corresponding code to apply to correct entity bean methods from my session bean.
Why not manage this logic in the data control framework by declaring which methods are to use as finder to get the data and which methods are to used to manage the persistence (persist, merge, remove) at object level.

Answer 1

Exactly what i mean
Actually Data Control architecture seem to be business object oriented with native method called from the same object to retrieve (the most part done)/synchronize the data with the database.
This is fine than your architectue don't provide a service layer like session facade.
With session facade you will access different business method to retrieve data and synchronize you data with the database not different method from the same object like application module do in adf bc.
I see two solution, the best is that you extend the data control framework to offer standard crud operation and a tools to map these with one or many methods in the same (has to be the same?this is a question of architecture, i think it has to be the same but the framework may have to permit to use an another one for flexibility) session bean. This may be the same from adf bc or top link but the method may be selected automatically by the tool (wizard).
The second solution is to build specialized sesssion bean with mandatory standardized methods and map this directly in the framework. The advantage i see is that the updates may be collected only once by the session bean by passing a specialized object manipulated by a specialized method.
In fact the best solution is something that combines mapping in the data control and a specialized set of methods in the session bean to execute a transaction in a whole at the server level.
What is to be decided if the framework (data control part) has to maintain a status for attribute (or set of attribute that correspond to a data control node) that are new inserted, updated or deleted and to synchronize only the changes.
I'm very surprising that something like that it is not already in the framework ?
How do I have to take care of changes made in a data control node object when i have used it to generate an adf faces by example in master / detail pages with an edit form like in otn examples. I see that the detail list in the detail table is updated so the data control is updated correctly by the edit form when i submit the changes but how to finally get this changes and synchronize with the database ?
ps i reply previously by mistake by email sorry ... not that the email is not exactly the same as the post here
Michel