on scalability
>> And even in that case, you end up calling metods, even if you call object "handles". <<
Taking credit for handles is going al gOOre overboard. A handle is simply an ID that points to a larger thingy. That is as much a database concept as it is OOP.
>> On the other hand, in a large project there happens a point where your own data structures/code become as complex as underlying DB access code/data. <<
IMO, the p/r model is *more* scalable for these reasons (at least):
1. One focuses generally on one task at a time, regardless of the project size.
2. Tables and large ER setups are easier to grok than large class networks. (Usually a lower protocol-coupling ratio also).
3. P/r tends to factor common patterns into tables instead of code structures, where non-programmers or newbies can manage the control info without ever touching code. The Report Manager example is a case in point: most new reports are simply specified with data in tables. OTOH, most OOP puts all that info in code, where attributes and algorithms are all mixed together in a hard-to-navigate soup.
ADMIN: is there an area we can take these battles so that they don't "clutter" the regular stuff? I know that some of you don't like these kind of discussions. Thus, if they can be moved to a dark corner somewhere, they may not bother you as much.
________________
oop.ismad.com