Re: Bill of materials / Groups in Groups

Kyle Lahnakoski wrote in message <387E452D.C813E8D2_at_arcavia.com>...
>I would say that a constraint rejecting an input to the DB is an
>anomaly.

Well, Mr. Celko's model is a particularly extreme example of this sort of "anomaly". Virtually _any_ attempt to insert a row into the database would be rejected in his model. The only way to modify the database is to use a stored procedure that performs the requested modification and then updates all the lft and rgt attributes. (The database travels through many invalid states on its way to becoming valid again. That's fine with me. I don't know if you would consider it an "anomaly".)

My point is that I can use the same approach to "normalize" my bad Employee/Department example. I've been trying to figure out what notion of "normalization" Mr. Celko was referring to when he made the comment about changing Chuck's id number. Apparently not any standard notion based on FDs, since even the original example (using employee names instead of ID numbers) was perfectly well normalized in that sense. He proposed another definition of normalization, and I tried to point out that that definition is in a certain sense much weaker than any of the standard definitions and that he probably didn't really mean to use it.

He may actually be using "normalized" as an authoritative-sounding synonym for "well-designed"; in that case I would ask him to consider using a different term which would more accurately reflect the subjective nature of the judgment. If there really is a _mathematical_ notion of normalization under which his model is better normalized than the standard adjacency model, I'm eager to know what notion it is. Received on Fri Jan 14 2000 - 00:00:00 CET