Best practices for model driven development using LiveCycle Data Services

What are your advises on using model driven development in developing enterprise applications. Adobe's LiveCycle Data Services looks very promising, I have found numerous tutorials/videos that shows how fast an application can be build by having methods/functions auto-generated.

What are the best-practices, is it good/bad to use those au开发者_Go百科to-generated methods, they can really save a lot of time.

All suggestions are welcome, also if you know some existing blog/discussion please let me know.

Make sure your model is very tight, and that your object graph doesn't bring down the whole database when all you are really trying to do is fill a combo box. ;) On our last Flex project part of our team sold our management on using Hibernate as the persistence layer on our legacy relational database. Part of the sales pitch was that the model could just be generated by a wizard. Well, maybe LCDS handles lazy loading better than BlazeDS, but the first time we tried to load a combobox, we pulled several megs of data due to the autogenerated relations, sometimes recursive, that the wizards had generated. These probably wouldn't have ever come to light in the course of a normal web app, due to lazy loading, but serialization will show you the exact meaning of "N + 1 problem"!!

;)