Re: Database design, Keys and some other things

You're right, my wording is sloppy (I don't mean identifier, I mean reference, and when i say a tuple's values defines it in RM, I am inferring this from its values being the only way to access and refer to it). I also don't mean to be overly critical of the RM - after all through the use of surrogate keys, you can achieve the desired effect.

I maintain the overall point though, in that any tuple (which after all is representing a concept, the finite partial map, which in mathematical notation would happily have a denotation) must have a reference to it available, as its existence cannot be referred to by a composite of its values alone (see the 20 year old car example above).

Yet any such reference is not an attribute such as the car's colour or engine size are - it's a way of denoting a value-less concept, representing the tuple as a whole. Given this is valid for any tuple, the data model imo should be handling this for you, not forcing you to crowbar a solution by fudging in user-generated artificial keys. Received on Mon Sep 26 2005 - 20:03:17 CEST