Re: Relational and multivalue databases

"Mikito Harakiri" <mikharakiri_at_iahu.com> wrote in message news:gqyZb.36$w.284_at_news.oracle.com...
> "Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message
> news:c16836$jht$1_at_news.netins.net...
> > Yes, it is simply a key-value -- that is function(key) = value type of
> data
> > storage. In fact, it makes more sense to call it a file structure than
a
> > database. When I first saw PICK (as a manager at a new place of
> employment)
> > I remarked "that is NOT a database"! I don't care what we call it, but
I
> > wouldn't want to go back to the DBMS's that my teams and I had worked
with
> > in the past. It would simply not be stewardly (in terms of dollars and
> > people time) to do so. But I'll keep working at determinig whether this
> is
> > a fluke or whether there really is a good logical reason for
> non-relational
> > data storage to be advisable in more instances than, perhaps, a
relational
> > model is appropriate.
>
> Dawn,
>
> Relational is not about the cost. It is not about storage either. Some
folks
> say it's about data management, but I would disagree. It's just a high
level
> programming model. Unless, you demonstrate some innovative Pick methods in
> that area, you'll have hard time finding good listeners here on cdt.
>
> Speaking about cheap solutions, there are open source databases...
>
Mikito -- I might not have been clear about the theory side on this. Here is my dillemma -- I have studied database theory and I have used many databases and data storage approaches. It seemed to me that there are many folks who have been taught or who believe that the relational data model actually leads to a better solution in data quality, database maintenance over time, etc. That is, it seemed from my studies that a company would be the best steward of their financial resources if they were to employ a relational database.

However, my experience tells me that the implementations of the relational model "seem to" (I admit I have no concrete proof of this) be more costly, without corresponding benefits, to the corporate owner.

If the relational model is not intended to yield a better solution, when taking into consideration all factors, than a non-relational model, then I could care less about it -- would you still care about it then?

So, let's look at the big picture of requirements for an application that includes data storage -- if we look at the overall cost of ownership (including data quality, ongoing support costs etc) of an RDBMS is it lower or higher than the implementations of other models such as PICK. My hypothesis is that it is more expensive (often considerably more) to employ an RDBMS. Is this irrelevant? I don't think so -- I think it tosses into question what the purpose of the theory is in the first place.

So, what is our goal in having a good theory of how to store and retrieve data? --dawn Received on Sat Feb 21 2004 - 02:37:38 CET