Re: Nearest Common Ancestor Report (XDb1's $1000 Challenge)
Date: Tue, 1 Jun 2004 08:58:56 -0400
Message-ID: <Q_KdnSYJ5Yui4CHd4p2dnA_at_comcast.com>
"Lee Fesperman" <firstsql_at_ix.netcom.com> wrote in message
news:40BBD212.70E9_at_ix.netcom.com...
> It appalls me how many are taken in by snake oil salesman (thanks for the
I agree with you, as far as you've taken it. But I think it goes deeper
than this.
quote, Bob
> Badour) like Neo who claim to have a better solution than the relational
model.
I think that most of the snake oil salesmen show up with a product, rather
than an alternate theory. I think this is very, very significant.
It's almost as though those who are looking for the silver bullet want
something that can be described thus: "we don't know why it works, but we
know that it does."
Now I'm one of those "practical people" that are viewed as barbarians by the
high priests of this forum. But this time, I'm going to say a few words in
favor of a model with a sound theoretical base. One of the purposes of a
model is to make visible the consequences of a design decision without
entailing the actual costs of implementing the design.
A model without a sound theoretical base is often capable of giving
misleading information about a proposed design. This can lend support for
a bad design, or draw support away from a good design. In other words, it
can lead away from success.
A model with a sound theoretical base can also do this, but it's easier to
understand the model itself, and to know its scope.
The most radical departure that Codd made in 1970 was to try to piece
together a theory of data structure whose internal consistency could be
proven. In other threads, I've been somewhat dismissive of "tautology".
But if I have to pick between a tautological system and a self contradictory
system, in the absence of very strong evidence either way, I'm going to go
with the tautological system.
It's the only practical thing to do.