Re: MultiValue Databases

From: Neo <>
Date: 13 Jun 2005 06:57:44 -0700
Message-ID: <>

> ... you do not have a clear understanding of the RM.
> If your problem domain includes several people with the same name,
> and you need to distinguish between them, you need to introduce
> an additional attribute, yes.

I hadn't considered that situation.

> What "integrity between multiple 222-2222s"? There is only one 222-2222.
> It is a value; it may be used all over the place, but it is still the
> same value. What kind of corruption are you afraid of? What management
> are you requesting? You are being very unclear.

I didn't realize the correct methodology for values. I think I am confused.

> Please. You can't tell a computer "minimize data corruption and maximize
> data management". Give me rules. Be explicit. Be formal.

Yes, I must admit that is quite imprecise and quite useless.

> > Each thing is represented once ...
> You call *that* integrity? That makes it hard to discuss it.

Le me re-think my thoughts.

> > how to use use RFD to store/query the data in above example?
> triplets might look like this
> Person type Class
> phone type Property
> john type Person
> john phone 111-1111
> john phone 222-2222
> mary type Class
> mary phone 222-2222
> bob type Person
> It's flexible, you can make your own "schema", and the "schema" is part
> of the "database" (for better or for worse). But you have no integrity
> (there are inference rules, but they are undecidable in the general
> case), and it is of course a network: pointer chasing, graph theory and
> record-at-a-time processing.

Interesting, I hadn't considered the pro's and con's of RFD type data models. Thanks for your input. Received on Mon Jun 13 2005 - 15:57:44 CEST

Original text of this message