Well, it's the second question of a pair about normalisation theory; the first will appear in about a week's time. (Steve's ways are sometimes hard to follow - but I'm not complaining, I saw that they would appear in the wrong order but didn't get around to asking him to change it, so it's my fault.)

My idea was to put a couple of questions that would go well with the fourth article on my series about normalisation. This is somewhat screwed up by the fact that I wrote an appallingly bad draft of the fourth article (so bad that Steve hasn't yet commented on it, probably because he would be embarrassed to tell me how bad it is) and haven't yet sorted it out, so the questions are appearing before the article insetad of after it.

It is really worrying that people don't unsterstand the representation principle (the base of all early work on normalisation - everything up to and including third normal form) and how normalisation adheres to it as far as elementary key normal form but abandoned it in Boyce-Codd normal form (and all higher normal forms) so that for many situations it is better to have a schema that has some tables not in BCNF or any higher normal form although most of the tables should conform to higher normal forms, because this lack of understanding of basic normalisation principles encourages schema designs that accomodate coding bugs with no error messages that with proper schema design would violate declared constraints.