Why not just dump relational?

Posted Jul 24, 2008 7:12 UTC (Thu) by jamesh (guest, #1159)
In reply to: Why not just dump relational? by Wol
Parent article: Drizzle: a lighter MySQL

While Newton's model might not have matched up with reality at extremes, it provides a useful
model for a large set of problems.  There are certainly cases where it will give the wrong
answers, but that doesn't invalidate its usefulness in other areas.

Similarly there are many problems that can usefully be expressed in the relational model,
which accounts for the success of SQL databases.  This isn't to say that it is a perfect fit
for all problems (I think I've heard the UniVerse example before), but it does have its uses.

Why not just dump relational?

Posted Jul 26, 2008 0:08 UTC (Sat) by Wol (subscriber, #4433) [Link]

I'd agree with you.

And I actually use relational theory to design my MultiValue databases. But just as Newton's
theories don't work on the very small, or very large, scale, so I find relational theory
doesn't work on the large scale (namely modelling any half-way complex system!). It's great
for modelling attributes of a single entity, and also the links between entities, but it fails
badly when imposed on a complete system.

"All data comes in rows and columns" - that's rule 1 I think. NO IT DOESN'T.

"All data is equal" - not sure which rule that is, but it immediately contradicts itself by
dividing data into relationships, keys (primary and otherwise), and attributes.

I could go on ... suffice it to say, I consider relational theory, as implemented in most
products, to be in serious breach of Einsteins' maxim - "make things as simple as possible BUT
NO SIMPLER". The relational database has been simplified too far - to the extent that all the
supporting stuff around it has to be over-complex in order to compensate.

Cheers,
Wol