Re: Unknown SQL

> >> No to all of the above
>
> Bob> So, you admit that object languages expose physical implementation details
> Bob> to the user...?
>
> no, I don't; as I've tried to say before, there may be ways of making
> references portable,

I never asked whether one can contrive kludges for transport. I asked whether the OODBMS exposes physical implementation details.

 e.g. in the way that CORBA IORs do. What exactly do you
> mean by 'physical' ? By physical, I understand 'hardware address'.

By physical, I mean all of the structures representing or represented by hardware addresses. For instance, a B+tree index is a physical structure. The physical structure increases the amount of memory required, increases physical data redundancy, reduces the number of physical page accesses for certain operations while increasing the number of physical page accesses for others.

A decision to cluster the data from one relation with the data from another relation -- ie. to ensure physical proximity on disk -- is a physical decision. A relational system does not expose such a decision in the presentation of the data, whereas an OODBMS does. The OODBMS will represent this as some form of containment relationship different from other types of relationships.

This is especially true of a product like Carl's that maps the Java classes directly to physical storage structures.

> Bob> Well, how one chooses to implement an application is independent of the
> Bob> DBMS. If the developers decide to turn a powerful functional, declarative
> Bob> DBMS into a navigational record processor, that's their problem.
>
> This debate is gonna spill over into a OO-war, it seems :-)

What do you mean? I am an object oriented programmer and have been using the technology for 14 years. Who am I going to war with?

> Bob> If neither Lisp nor Java exposes physical implementation details, then I
> Bob> must assume that Lisp supports object references,
>
> yes.
>
> Bob> always passes objects
> Bob> by reference,
>
> yes, more or less.
>
> Bob> supports single inheritance of implementation and multiple inheritance
> Bob> of interface?
>
> No, multiple inheritence and multiple dispatch of implementation (interfaces
> don't exist).

In other words, the two different languages expose different physical inheritance and delegation mechanisms. Ergo, the languages expose physical implementation details to the user.

Regards,
Bob Received on Tue Jun 05 2001 - 01:28:51 CEST