Re: The wonderful world of keys

Thanks for all the replies. I find this sort of stuff particularly helpful, and it has really started to help me distinguish a database with an inference engine (such as good ol' prolog for example). It has become clearer to me that inference engines record propositions, whereas Databases /model/ propositions. The first allows inference (often useful for some tasks). But databases provide a different service - by raising propositions to first order players, with their own properties and types, they allow one to manipulate large information sets in powerful, robust and unforeseen ways.

It used to concern me slightly that the externalization of a predicate meant that I lost the ability to infer /within the system/ - a database cares not whether two attributes are related by a logical 'and' or a logical 'or' for example, and relies on you to determine what a Join will produce. However Material Implication is different, as it affects all sorts of collection anomalies as information changes, and its good to know that this is not just some RM artifact, but part of the nature of how a proposition is predicated. My candidate keys no longer look simply as though they are just 'proposition identifiers'; I am enjoying the obedient hum as they do their work, knowing there is something more substantial to them than I had realised. Even BCNF now has a warmer glow, part of the nature of information rather than just some old rule in a database model. But hey, maybe that's just the caffeine overdose again. Received on Sat Jan 27 2007 - 03:21:41 CET