I agree with Ajith in principle. But at the same time I would also like to add that CMP generated queries, though supported, through lot of indirections, by people who know SQL well, ultimately has to serve a generic framework. And this is where the inefficiences pour in. If you try to look at what the container does to a possible simple BMP query, by turning on logs, you would know what I mean.
But ofcourse, ultimately the resulting performance requirement and time to be spent coding makes the difference. And it's here that 2.0 CMP is said to overall score well.
Sridhar, I haven't ever tried to understand the 1.1 CMP model and hence wouldn't know the changes in 2.0 specs. Probably, somebody who has followed ejb spec journey from its initiation can comment. Anyone ?