Re: More on identifiers

From: David BL <>
Date: Sat, 6 Jun 2009 06:36:43 -0700 (PDT)
Message-ID: <>

On Jun 6, 7:46 pm, "Walter Mitty" <> wrote:
> Do you remember Neo and his attempt to come up with a database (a language,
> really)
> that could capture all facts, no matter how they were structured?
> Are we going down this same road again?

Around the time I first arrived at the group, someone called Neo posted occasionally, but he was largely ignored. To be honest I never took the time to read his posts so couldn't comment.

In some respects my suggestion is only incremental. The RM is applied like normal, except there is the option of no longer unnecessarily labelling things by using DVAs.

Note that the RM as described by C.Date doesn't make any prohibitions on a domain type. To that extent my suggestion of using DVAs is fully consistent with the RM.

When a DVA is to be used, I'm suggesting the normal practise of harnessing the RM to record information using propositions associated with tuples in relations. Note for example that one develops a normalised relational schema as per usual. The only difference is the narrow focus or context. All propositions are interpreted in the context of some particular thing that needs to be identified by its recorded properties.

In that sense I think the idea of a DVA is easier to understand than an RVA - at least in terms of a clean conceptual basis. I think it's a more consistent use of the RM, and better accounts for the implicit context in which a database value is intended to be interpreted. Remember that a database variable always exists in time and space and is meant to be interpreted in a corresponding, implicit context.

The part that's novel is just the idea to use a relational database value as an identifier. Received on Sat Jun 06 2009 - 15:36:43 CEST

Original text of this message