Re: Relational vs network vs hierarchic databases

"Dan" <guntermann_at_verizon.com> ...
>
> "Ja Lar" <jalar_at_nomail.com> ...

<snip>
> How do you talk about a hierarchical/network model (when you are really
> talking about systems) without talking about physical/implementation
> matters. That was the primary basis of criticism for why the relational
> model was superior (remember the term physical indepedence).

The criticism was the other way around: that network/hierarchic databases was better (the claim by the "OO people"). Such a claim is not substantiated by anything (in general, of course).

>> In this case, you might be correct in one or more concrete cases. (But
>> you
>> might be wrong as well).

> I'm not sure what you mean? "You might be right, but you might be wrong?"
> Care to be more concrete?
If you are wrong or not in a concrete case depends on, well, on the concrete case (actual DBMS, actual problem/domain, actual use). It is not relational or not per se that determines the performance.

>> Point is that "relational" and "performance" are orthogonal - and thus
>> not
>> contradicting or compeeting in any case.
>
> What does orthogonal mean? Does it mean that the intersecting vectors of
> relational and performance never ever really intersect? No.
Relational is logical and performance is physical, so they don't intersect (and maybe I should not even call them orthogonal).

> Yes. The relational model is a logical model that is orthogonal to
> "performance" in the abstract. But if you are going to compare it to a
> hierarchical or network model and criticise those models
I has nowhere criticised hierachical/network models (yet...)

>for being constrained and hindered because of the use and dependence on
>pointers, which are physical,
hierarchical /network models are not physical, but there implementations are of course

>then their is no reason not to recognize their strengths in terms of their
>physical manifestion as well.

(Almost) Any type of "physical manifestation" can outperform another type in a _specific_ context. This is independent of model used. Eg., if I have a set of data I never want to change, and just one report (query) I exclusively and repeatedly use, I could chose to represent this "database" in a flat file, containing my report. Outperforms anything. But as soon as I need to update data, or do a new query not foreseen, I'm confident that a physical implementation of a relational model will serve me best... Received on Tue Nov 09 2004 - 18:19:18 CET