Re: Unknown SQL

>> >- Performance of object databases is clearly by far superior to
 relational
>> >databases for certain tasks.
>>
>> Wrong. Performance is a function of the physical structures used to store
>> data. The relational data model makes no restrictions on these physical
>> structures. As such, it can certainly achieve the same performance as any
>> other data model.
>
>For *certain tasks* I can prove the performance advantages empirically. Why
>bother about your theory?

No, you cannot. For certain tasks, you can demonstrate a performance advantage over specific products configured in certain ways. You cannot prove that no relational solution can exist with equal performance unless you test against all possible relational solutions. Since all possible relational solutions is an unlimited set, you cannot complete the task.

In any case, if you understood the distinction between logical and physical, you would know better in the first place.

>- Relational databases need to split single objects up on multiple tables.

No, they do not. If you made any effort to educate yourself, at all, you would know better. I suggest you familiarize yourself with Date's and Darwen's _Third Manifesto..._ or Fabian Pascal's most recent contribution. I don't remember the exact title off the top of my head.

>Passing keys back and forth to link the pieces together simply costs
>performance.

This is not necessary, which makes your point moot.

>- Object databases can analyze objects directly, without the need to
 convert
>them to a SQL string representation. That is much faster.

Bullshit.

>- Many usecases for commercially used relational databases involve
>incredible driver overhead. Converting data over JDBC -> ODBC for instance,
>drastically slows down performance.

Criticisms of JDBC and ODBC are hardly valid criticisms of relational databases. They do not equate.

Regards,
Bob Received on Sun Jul 22 2001 - 01:28:35 CEST