Regarding where to put logic, it's really best not to be dogmatic about this. Putting logic in the db can have some security advantages but it has other security gotchas (for example, SQL injection inside a stored procedure may be possible in some environments). Similarly it can have some portability advantages and disadvantages.
The real question is that you want to ask what you want to be portable. If you put it in the db, it becomes easier to integrate programs written in different development environments with a minimum of security gotchas. If you put it in the application, then you have to create the interfaces on that level in middleware. Both approaches are doable. On the other hand if you are writing software you want to be deployable in MS SQL shops and Oracle shops, then you have to write portable SQL (avoiding gotchas) and put the logic in the application.
When we started rewriting the LedgerSMB codebase, we decided to move logic into the db because this would make it easier to retrofit security onto a very insecure codebase (by not trusting the application), and because we wanted to support languages other than Perl. A few years later this is bearing fruit and indeed I am reading the manual because I am writing integration libraries in PHP. I am not much of a PHP guy (I haven't programmed PHP since PHP 4.x) but I can write modules that let folks write code to interface securely with LedgerSMB's database logic. You can think of the stored procedures as being named queries which are shared between applications, or as API's to the database. But again, this approach is not universally applicable.
If portability between db's is not a major requirement, then I think the best approach is to do as much as possible in SQL queries and put those in the database. A couple hundred lines of SQL can replace a couple thousand lines in Perl or PHP, and is fundamentally easier to debug. Of course that won't work for everyone.
