Re: Composite attributes

>> This has been discussed recently in Database Debunkings -
http://www.dbdebunk.com/ - and not in wholly vitriolic terms, either. <<

I was slightly misquoted in that article. In my book, DATA & DATABASES, I made the point that people try to "fake" repeated groups and arrays in SQL by using numbered columns that are NULL-able (i.e. child_1, child_2, ..) which leads to a bad design. In the next edition, I want to add the current trick of putting a comma separated list into a VARCHAR(n) column (i.e. 'Tom, Mark, Larry') and then extracting values from it.

I would say that the value has to be scalar or if it has "parts", the parts have no meaning by themselves in isolation. Something like temperature is easy -- one value, no parts.

But an international tire (tyre) code is one unit, even tho it is made up of diameter, material and width (i.e. 155SR15 is a 15 inch diameter, steel radial that is 155 cm wide). A 15 inch diameter cannot exist by itself -- it has to be the diameter of something.

Dates are tricky. A date is a scalar, even tho we can extract month, day and year from it. However, it is really a shorthand for a duration that lasts 24 hours. Now we are into discrete sets versus the continuum and all those problems ... Received on Wed Dec 11 2002 - 21:15:14 CET