Re: By The Dawn's Normal Light

Dawn M. Wolthuis wrote:
> ABSOLUTELY!! OK, so who in the database world thinks that there is anything
> useful in talking about data in 1NF? Are we ready to ditch it completely
> and talk about data in 2NF & 3NF without requiring them in the meaningless
> 1NF first? --dawn

I think it's still a useful concept.

But I think rather than saying "this relation is in 1NF" it makes more sense to say "this DBMS is 1NF". Which kind of makes it a bit of a misnomer I guess.

By 1NF, you're just saying that in this DBMS, tables are sets (no duplicates) and values are atomic (the relational engine can't look at the internal structure of values).

I think that when talking about RDBMSs 1NF is useless, because relations here are automatically in normal form.

But if you are talking about some kind of DBMS that allow the DBMS to delve into relations nested inside relations, then it does have relevance. Maybe then you could then say that a relation with no nested relations (or at least none whose inner values are visible to the relational engine) is in 1NF. But one with such relation-values isn't.

For a SQL DBMS maybe you could say a table is in 1NF iff it has a primary key (or at least a candidate key) defined.

I think conceptually as well the "repeating groups" idea of 1NF is useful for users who are used to the spreadsheet way of storing data, where the user can add extra columns as easily as they can add rows. So really it's just expressing the idea that in a DBMS, columns are difficult to add, rows are easy to add.

Paul. Received on Tue Oct 26 2004 - 22:40:12 CEST