Re: 1 NF

On Mar 2, 3:35 pm, "Marshall" <marshall.spi..._at_gmail.com> wrote:
> On Mar 2, 4:54 am, "JOG" <j..._at_cs.nott.ac.uk> wrote:
>
>
>
> > * Using a mathematical relation means that all the propositions
> > collected in that set have to have the exact same structure.
>
> Lately playing with ideas about constraint systems I've
> been toying with the idea that we could dispose of
> types as a *requirement* and make them optional,
> the same way that constraints are optional. However
> on further examination, it became clear that "having
> the same structure" was not a requirement that could
> be removed. In other words, we can get rid of column
> types and still have well-defined semantics, but we
> must retain the idea of a relation type, and that type
> must constrain what attributes the elements of the
> relation have, even if it doesn't constraint the values
> that elements' attributes have.
>
> This seems very much like what you're saying.
>
> Marshall

Exactly. I have done similar exploration, and the idea of the collecting set type (read relation type for RM) seems essential. I'm currently looking at different ways of predicating those sets and its certainly giving me interesting food for thought, especially in the area of views. Nowhere near thoroughly analysed yet though, so please consider this wanton hand waving. However my instinct is that if we are going to make progress in the issues of view updatability and missing information it will be via these sort of areas. Received on Fri Mar 02 2007 - 18:54:17 CET