The Data Model Resource Book, Volume 3: Universal Patterns for Data Modeling

Silverston c05.tex V2 - 11/21/2008 3:04am Page 195

Level 1 Classification Pattern 195

Thus, we want to emphasize that we would almost always not recommend to use this

pattern (or for that matter, most of the level 1 patterns) as a basis for any long-term or

significant implementation because there are great pitfalls that exist in implementing

this pattern, such as redundant and inconsistent data. However, instead of having

to show a ‘‘normalized’’ data model to a business representative (who doesn’t

care if the model is normalized or not and only cares that you have captured

their needs), we have found that this type of pattern can be an effective way to

illustrate data requirements to communicate with business representatives in

order to better understand their needs.

We have emphasized not to use this pattern for implementations; however,

there may be rare cases where the classifications and instances of classifications

are completely unchanging or static and where redundant data does not cause

major issues. You could make an argument that if the classification required

only a simple attribute, you knew that the entity would only ever have

this one classification, and the physical implementation benefited from a

non-normalized design, this may be a possible implementation. For example,

perhaps it could be a possibility to use a gender type attribute for PERSON

where the values are and always will be ...