Advantages JCA DB Adapter vs EJB

Question

The recent acquiring of Bea by Oracle unveiled a bunch of JCA adapters to do connectivity to (backend) resources. One of them is the DB Adapter. A wizard in JDeveloper creates a JCA adapter that internally uses Toplink. Although this looks like a nice approach still I have some hestitation to use it. You could decide to do the interaction with the database based on EJBs (or other more direct approach). No one in our team has any experience with JCA what so ever. So when the wizard or generated adapter fails it will be quit a problem for us.
The wizard based approach of the adapter gives you quick results. The downside is the restriction of what the wizard offers. If it does not fit you can edit the toplink stuff. But this would mean you have to dive deep inside the JCA internals. From a wizard to that is quit a big "jump". The need for this OR mapping somewhat puzzles me anyways although I can image that building a configurable adapter based on such a framework makes life easier (It is always good to combine existring stuff into a new product than building it all from scratch).
If this is the way forward into connecting with a database from OSB (in our case) it might be needed to invest in this approach. Questions are raised why we cannot use an approach that does not introduce a new stack of technologies (JCA, Toplink). We will be connecting with a single big Oracle database/Schema. Currently we are doing simple queries or calling PL/SQL functions. So looking at the current requirements the wizard will be able to handle that.
Does anyone has an opinion in this or some guidelines/best practices.
Regards
Frank (ALSB/J2EE background)

Answer 1

The downside is the restriction of what the wizard offers. If it does not fit you can edit the toplink stuff. But this would mean you have to dive deep inside the JCA internals.If it does not fit you can edit the toplink stuff. But this would mean you have to dive deep inside the JCA internals. From a wizard to that is quit a big "jump". The need for this OR mapping somewhat puzzles me anyways although I can image that building a configurable adapter based on such a framework makes life easier (It is always good to combine existring stuff into a new product than building it all from scratch).The value addition is highest when use-cases require connecting to various DB instances (eg Oracle, Sybase,DB2 etc). In such cases DB adapter will allow you to develop application very fast, with out spending much time in understanding and coding for each DB. Other value addition of using DB adapter is it allows getting inbound events from DB and has support for multiple strategies. As with any other technologies, improvements in productivity are accomplished by constantly adding technology stacks. So toplink/JCA is one such stack which improves productivity for accessing DB. There might be some rare cases which DB adapter doesn't support. In such cases we can always go with JavaCallout/JDBC approach.
Currently we are doing simple queries or calling PL/SQL functions. So looking at the current requirements the wizard will be able to handle thatDB adapter should well suite this case.Only issue you have raised is new stack being introduced which needs mastering.
These productivity improvements change from case to case. Eg. A use case that will only require requiring only one simple insert into Oracle DB is better coded with JavaCallout/JDBC rather than using DB adapter. A use-case that requires that we poll on table and get the update on table is very complex to implement by code(ie we have to write the logic for the polling /avoiding duplicate records) . In such cases it makes sense to use DB adapter rather than write java code.
At the end it boils down to deciding if the new stack is worth adopting given the productivity improvements it can bring.
My two cents
Manoj