Re: Proposal: 6NF
From: David Portas <REMOVE_BEFORE_REPLYING_dportas_at_acm.org>
Date: 30 Sep 2006 10:45:03 -0700
Message-ID: <1159638303.513871.92440_at_m73g2000cwd.googlegroups.com>
> There is a problem with your reasoning. The aim isn't simply to record
> propositions of fact: it is also to make use of those facts. Recording
> propositions is only about 5% of the typical activity for a database. With
> today's technology, querying multiple tables is almost always more expensive
> than querying one, and a database in 6NF would force most queries to involve
> multiple tables. Even if you separated only the NULLable columns into
> different tables, the effort required to perform many simple queries can
> become impractical. For example, if you have a table with 30 columns where
> 15 are NULLable, you would need as many as 16 tables to eliminate NULLs. As
> a consequence, every query that would have involved a NULLable column would
> now require a join.
Date: 30 Sep 2006 10:45:03 -0700
Message-ID: <1159638303.513871.92440_at_m73g2000cwd.googlegroups.com>
Brian Selzer wrote:
>
> There is a problem with your reasoning. The aim isn't simply to record
> propositions of fact: it is also to make use of those facts. Recording
> propositions is only about 5% of the typical activity for a database. With
> today's technology, querying multiple tables is almost always more expensive
> than querying one, and a database in 6NF would force most queries to involve
> multiple tables. Even if you separated only the NULLable columns into
> different tables, the effort required to perform many simple queries can
> become impractical. For example, if you have a table with 30 columns where
> 15 are NULLable, you would need as many as 16 tables to eliminate NULLs. As
> a consequence, every query that would have involved a NULLable column would
> now require a join.
Why does the state of "today's technology" indicate a problem with JOG's reasoning? If Codd had only concerned himself with the technology of his day then we wouldn't have a relational model at all.
> NULLs are as much a theoretical nonsense as the empty set or the imaginary
> component of a complex number.
Why do you say that? Imaginary numbers and empty sets are real values with the normal properties that mathematics and logic expects values to have. Nulls are not. As Date has pointed out, the "Assignment Principle" is so fundamental that it wasn't even thought worth mentioning - until the designers of SQL broke it!
-- David PortasReceived on Sat Sep 30 2006 - 19:45:03 CEST