Re: DB clasical structure violation

On Thu, 4 Jul 2002 23:38:57 +0100, "Anthony W. Youngman" <thewolery_at_nospam.demon.co.uk> wrote:

>Basically, chaptered rowsets (as I believe MS call it), or "nested
>tables" as it's simpler to refer as.

This 'feature' is also in the relational model. You can have relation typed attributes and be still in 1st normal form.

>This breaches 3NF, but as you can see, I don't have any redundant data
>so I don't have integrity problems there.

This does not break 3NF. If you don't have redundant data and integrity problems then your data is normalized.

>And one only has to look at MS's "the database is the file system"
>(where they are planning to use SQL-Server as the disk format) to say
>"Pick was there 30 years ago".

This may be true, but SQL server is not a relational DBMS.

The question is if a true relational DBMS would be better than Pick, and I think the answer is yes.  

>Don't forget I said the Pick language is a 3.5GL. So you're a lot
>"closer to the metal" than you are with the abstractions of SQL.

The trend of the developement history is "going far to the electrons" :-), and raising the level of abstraction.

> The
>only sort of file Pick has ever used is the hash file. And for the last
>twenty years, I've never had the need to manually rehash a file - the
>system does it for me, on the fly, and so fast I don't realise it's
>doing it. MS and Oracle are now copying our techniques for the internal
>management of their data inside the database.

There are nothing wrong in copying good ideas for the physical layer.

Alfredo Received on Sat Jul 06 2002 - 23:00:33 CEST