Re: Object identity


Gabriel Claramunt wrote:
I haven't either, but if you read the recent thread called "OO versus
RDB" you will see it.

I started, but I get bored quickly with holy wars, for me OO works better
than RDB, but I don't loose sleep if somebody prefers RDB: both can be
implemented with a turing machine, and with both I can implement a turing
machine...

IMO the holy wars relate to the two extremist camps that each think
that their formalism is right for every problem. They come up with
lots of useless, generic hand waving arguments.

I'm not in either camp. I instead see that both formalisms are
important and useful, and they basically solve different types of
problems. I wanted to understand what problems need RM and what need
OO. For me this is practical and useful. IMO the nature of object
identity is a crucial factor in this. If you want to store knowledge
about entities (as is so common in business applications) then use RM
(for that aspect of the total solution). If you want to create objects
that semantically need independent identity from entities, then prefer
OO.

I think the OO community needs to spend a lot more time properly
defining concepts like object identity, value-types versus
object-types etc. The discussions on comp.object reveal disagreement
over the fundamentals. I haven't seen a book on OO that is anywhere
near sufficiently formal.

Some proper science should stop all that silly rabbiting on.

Cheers,
David Barrett-Lennard

.