Re: Relational and multivalue databases

"Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message news:c107c7$q5i$1_at_news.netins.net...
> Wow -- thanks, Eric! I appreciate your responses.
>
> I'm sure you'll understand if it takes me a little time to address your
> points. Also, you are responding to a dialolg from 2002 when I was
starting
> to delve further into researching why the practical experience I had did
not
> align with the database theory I had learned. I've learned a bit more in
> the past year (so I'm still capable of it ;-) and will accept several of
> your points without need to respond.

Not a problem, I know it's a huge posting.

> Along with the fact that I'm probably more interested in theories that
yield
> developer productivity than those which yield a more academic goal,

I'll posit that at least in shops where I've worked, provable correctness is far more than an academic goal. In fact, the lack of such is the biggest productivity deterrent I see, hands down. Its handmaidens are closure and referential transparency, to which relational caters nicely.

> the biggest areas of disagreement I have relate to
> 1. modeling propositions as relations rather than functions that
represent
> graphs (mathematically)

Please explain and/or give an example of such a function.

> 2. referential integrity as well as type constraints and where these
should
> be specified and enforced.

This one I'll probably argue about more - suffice it to say that declaring them once and generating the necessary enforcement (e.g. on the client) will enable higher productivity and ensurable correctness. Where do you propose specifying and enforcing them?

• erk