> The basic problem is that writing good, high-performing SQL is hard enough as it is. You have to understand indexes, clustering, execution plans, and query op

> The basic problem is that writing good, high-performing SQL is hard enough as it is. You have to understand indexes, clustering, execution plans, and query optimization to get good performance. Not to mention cache specifics (something that has personally bitten me). To be honest, my preference is actually to write a service locator against stored procedures. This allows queries to be maintained separately from the application (and loosely coupled from them also). This is not too far from what iBatis does except that I would use stored procedures instead of mapped sql. What db's are you using? If Pg only, interested in porting the PGObject approach to Java?