Re: candidate keys in abstract parent relations

From: Forrest L Norvell <spankysyourpal_at_gmail.com>
Date: 20 Jan 2006 15:13:14 -0800
Message-ID: <1137798794.849190.208720_at_g14g2000cwa.googlegroups.com>


David Cressey wrote:
> Maybe an "Album" isn't really an entity at all, but instead a "reified"
> relationship between tracks and releases.

That's exactly what it is. The album is a reified notion that allows me to imply that the "same" album can be put out as multiple releases, i.e. the same collection of tracks released multiple times with slightly different metainformation. The main thing I have left to resolve is how to deal with Release-specific metadata that reaches down into the Album and its dependencies (like run-out groove etchings, which vary from release to release).

Anyone who wants to understand what I'm up against should take a look at the KLF discography:

http://www.klf.de/discography/index.php3?search=

Lazlo Nibble's discography format is extremely flexible and subtle, and he's able to concisely capture in freeform text exactly what I'm trying to shove into my model. All of the ramifications of that format, though, get mind-bending in a hurry. Limited editions! Colored vinyl! Releases with identical titles and catalog numbers but different tracks! Albums with absolutely identical collections of tracks but variant track names!

A lot of music cataloging systems canonicalize this data, but record collectors are generally pretty fundamentalist: since they want every variant version, even if normal human beings would think two variants were absolutely identical, the text is all, and the "text" in this case brings with it a lot of metainformation that's invisible to casual music consumers. Received on Sat Jan 21 2006 - 00:13:14 CET

Original text of this message