Re: Can we solve this -- NFNF and non-1NF at Loggerheads

From: Dawn M. Wolthuis <dwolt_at_tincat-group.comREMOVE>
Date: Sun, 6 Feb 2005 14:31:04 -0600
Message-ID: <cu5uqe$9sf$1_at_news.netins.net>


"Dan" <guntermann_at_verizon.com> wrote in message news:K1kNd.6968$uc.2841_at_trnddc04...
>
> "Dawn M. Wolthuis" <dwolt_at_tincat-group.comREMOVE> wrote in message
> news:cu40to$ck1$1_at_news.netins.net...
>> "Alan" <not.me_at_rcn.com> wrote in message
>> news:36lbt1F50dlt4U1_at_individual.net...
>>>
>>> "Dawn M. Wolthuis" <dwolt_at_tincat-group.comREMOVE> wrote in message
>>> news:cu2llf$jq3$1_at_news.netins.net...
>>>> "Alan" <alan_at_erols.com> wrote in message
>>>> news:36htmvF54uqs0U1_at_individual.net...
>>>> <snip>

> Hi,
>
> I must admit that I am very uncomfortable with the situation of someone
> taking ownership and redefining an aspect of someone else's model. Date
> and Codd did not always agree on "the model". For example, see
> http://www.dbdebunk.com/page/page/1706814.htm.

Yes, agreed.

> For me personally, I recognize the possibilities of RVA's as an extended
> model (or alternative one) and I am open to it, but I still have some
> serious concerns about the implications of an RVA-based or NFNF relational
> model.

Makes perfect sense. I would be there with you if my experiences to date had been different.

> This does not say that such models are invalid, but rather that the model
> introduces some complexities

Yes, the model is more complex, but some things are simplified -- more on that sometime.

> and implications that I haven't fully worked out yet. Now, without making
> any presuppositions and speaking in the most general sense, more
> complexity for the same amount of expressibility and power is generally
> not a good trade-off.

Agreed. I have seen significant benefits in flexibility (maintainability) and hope to be able to put that in words at some point better than I have to date.

> Until that time where I can fully understand and appreciate the
> implications, I will rather continue to feel more comfortable with a model
> that adheres and comforms to a uniform and consistent expressiveness of
> first order logic

Again, if I had not experienced something different that seemed better from a TCO perspective, I would be in the same camp as you.

> without special "safety" rules being necessary, particularly in the domain
> of relational tuple calculus.

I don't have to speculate on whether this would work because it does, but there are tradeoffs.

> Date often speaks of models in terms of structure, manipulative aspects,
> and integrity (this is the one the produces the most concern for me with
> RVA's). Or rather, he regards these three items the essential components
> of model specification. I personally think that all three aspects are
> modified in some degree from Codd's original model with this new "1NF"
> (e.g. RVA, nest and unnest, and who knows what else with n-order
> traversals of nesting levels with foreign keys). Therefore, I think this
> extension is a model that is separate and discernable one, not one that
> "redefines" and supercedes the original.

Yes -- agreed!

> If I understand you Dawn, this is the point you are making. Don't
> redefine 1NF, call Date's model something else.

Yes. If there is now a new model that is not an easy migration from the state of relational theory, such as this change that really significantly redefines 1NF for practitioners, the theory could definitely have a new name and the old terms should be left in tact.

> Everyone won't agree with this particular opinion of course. I can accept
> that.

Your opinion is much appreciated! Thanks. --dawn

> - Dan
>
Received on Sun Feb 06 2005 - 21:31:04 CET

Original text of this message