Re: Clean Object Class Design -- What is it?

>Additionally most object databases do not work as nicely, if indices do not
>fit into RAM.
>
>Implementing file-only operation mode is more natural for a relational
>system that uses fixed length pages, tables, blocks, rows and columns.
>Relational systems already had to work smoothly before RAM was as cheap as
>today.
>
>Of course an object database implementation could also choose to use a
 fixed
>length structure for indices but it would be somewhat relational. :-)

Are you truly as ignorant as you present yourself? The relational data model imposes no restrictions on physical representation and has nothing to say about physical storage, yet you direct every one of your criticisms of the relational model at physical storage. That's like dismissing the utility of sky-scrapers because climbing the stairs takes too long.

The fact is most commercial SQL databases use variable length structures to represent table rows and use variable length compressed keys in their indexes.

>Hopefully we are heading to hybrid systems that provide the
>best-of-both-world functionality.

It is impossible to hybridize a relational dbms with a navigational dbms. As soon as the dbms requires navigation, it ceases to be a relational dbms.

>Query deficiencies of object databases definitely are *not* a problem of
 the
>theoretical possibilities.

Actually, they are a problem of the logical data model. How does forced navigation expand the "theoretical" possibilities?

>They are only due to smaller development time and
>budgets until today.

Are you saying that the developers of IMS had too small a budget and too little time to overcome the query deficiences of their navigational database model? Received on Sat Aug 04 2001 - 20:44:31 CEST