Re: Date, Darwen, Pascal and the alternative to Nulls in the RM

mountain man wrote:
> The only way to avoid nulls is to avoid change management.
> That's it really. End of story. Change management will eventually
> breed the occurrence of nulls.

In the relational model I agree. In fact I am more and more of the opinion that a null-less RM is implausible. If you can't handle dealing with the jarring nature of nulls, then you don't want RM.

There appears to my my mind a simple reason for this: you can eradicate nulls to your heart's content, decomposing one table after another and normalising-a-sweet-fandango . But at some point, someones going to ask a question of the database and its going to have to JOIN those tables, and back will pop those nulls. A query language has transitive closure over relations but, by its very design, these db-relations just have to be capable of having gaping holes in them.

I'm starting to believe Codd was most likely aware of these issues, and rather conceded to pragmatism. He realised something was afoot mathematically when he added column headers, which recent proponents tend to ignore, and similarly I imagine he knew that for the RM to function, nulls were going to be a necessary evil. Received on Thu Mar 23 2006 - 05:33:57 CET