Re: Changing Object Type in OODB

>> >> In another thread, a Java-only OODB proponent proposed the following
 as
 an
>> >> obvious, straightforward design not requiring any thought:
>> >>
>> >> >class Person
>> >> >class Employee extends Person
>> >> >class Manager extends Employee
>
>I think someone means me, here.
>Sorry for this simplified example that I usually use.
>Patrick has pointed our very well, that it might also be wrong.

If you knew it was a possibly wrong, simplified example, why did you try to claim that its obviousness made OODBs superior to relational DBs? Why did you criticize me and the relational model simply because I pointed out that one must know the full requirements prior to engaging in design?

Were you expecting some knee-jerk design you could poke holes in? Or were you just trying to mischaracterize the truth to sell your product?

>[changing an Employee object to be a Manager]
>
>Independant of my possible bad class design, the problem brought up could
 be
>of importance in practice. I agree that Java does not provide a means for a
>downcast which would be necessary here, in order to have other objects keep
>the same reference within the database.
>
>A feature
>DatabaseEngine.cast(Object, Class)
>could very well be very useful for object databases.
>Thank you for the excellent suggestion.

Are you saying that an object's type or class is not an essential property of the object? This seems to run counter to everything that object proponents espouse.

>> Not really but those things also occur in OODBS to a heavy degree. And
>> even to the worse you can´t that easily change your db scheme once you
>> have it nailed down inside the db.
>
>This prejudice about the rigidness of object databases is widely spread.

And well justified. All network model database schemas are rigid -- this happens when the logical view of the data is equivalent to the physical layout.

>Indeed it is well-founded by the difficulties most object databases have
>with schema changes.
>
>I am very positive that we have not seen the end of development here.

I suppose someday the OODB folks will admit to the necessity of physical independence. When they do so, they will realize that the logical model of OODBs provides no performance advantage, and that the data model is independent of performance. The question is: Will they realize that the relational model has been the solution all along?

Regards,
Bob Received on Mon Jun 04 2001 - 01:48:11 CEST