Re: Relational Model and Search Engines?

• Post for FREE via your newsreader at post.usenet.com ****

"Paul" <paul_at_test.com> wrote in message news:dfKlc.39446$Y%6.5362560_at_wards.force9.net...
> Leandro Guimaraes Faria Corsetti Dutra wrote:

> > The relational model is purely logical. Physically the
> > implementor is free to use whatever tricks he may wish.
>
> I've seen this argument used a lot and it doesn't seem quite right.
>
> Suppose I have two logical models that do the same thing. For
> implementing them physically, the best possible algorithm for one might
> be O(n^2) but for the other it might be O(1).

Well, if you already have another logical model that does the same thing for which an O(1) algorithm exist,
why you wouldn't use this O(1) algorithm instead of the O(n^2) one for implementing the first logical model ?

In other words: the situation you described is possible ? I can think of a single possible reason for this to happen: when the cost of translating from one model to another is O(n^2). In this case are you sure the two models "do the same thing" ?

> So what I'm saying is that a logical model is not totally independent of
> any physical implementation of it, there may be a upper bound on how
> efficient it can be.

This is true. But this does not require the existence of an equivalent more "efficient" model.

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

• Usenet.com - The #1 Usenet Newsgroup Service on The Planet! *** http://www.usenet.com Unlimited Download - 19 Seperate Servers - 90,000 groups - Uncensored -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=