Re: Natural keys vs Aritficial Keys

"Roy Hann" <specially_at_processed.almost.meat> wrote in message news:982dnVavb_XQM9DXnZ2dnUVZ8oGdnZ2d_at_pipex.net...

> My point is that the rows are logically duplicate but they have been
> rendered spuriously distinct by tacking on a meaningless but unique
> attribute. So such an update is easy (perhaps inevitable).

This is an enormously important point, one that gets overlooked time and time again in the discussion of databases.

I take this one step further, and construe the attribute that was tacked on to be a "logical address". If that sounds like an oxymoron, so be it. If all linkages between data stored in different rows is made by foreign key references to these "logical addresses", I claim that the fundamental data model is now a graph model of data, and no longer a relational model of data.

So I further claim that the people who pursue this design methodology end up acheiving nothing that could not have been acheived equally well or better by using a graph database instead of a relational database. And I think these results are used by a lot of people to spuriously disparage the real value of the relational model.

You might disagree with my add ons, but I want to reiterate that the point you made was an enormously important one. Received on Fri Jul 03 2009 - 09:19:23 CEST