After 30 Years,
The Traditional Normalization Procedure's Fallacies Are Exposed!
Most I.T. people know that to have an efficient,
internally-consistent relational database system, you need to start
with a normalized schema—and then denormalize, only if necessary, in a
deliberate, controlled way. What has only recently come to light is
that the traditional normalization procedure, based on "nonloss
decomposition", is insufficient to give you a Fifth Normal
Form (5NF) or even Fourth Normal Form (4NF) schema design.
The real way to do a fully-normalized schema design
has been around for a while - but it was never marketed, or even
recognized, as the only effective way to do schema normalization. But a
colleague, Terry Halpin, and I (who have both taught database design at
the university level for years) discovered not only that the
traditional normalization procedure was defective, but that the design
procedure we teach -- conceptual information-modeling (in the form of
Object-Role Modeling, ORM) -- is the only effective, practical way to
ensure a fully normalized schema.
We wrote a
on the topic, which we presented
at the 2008 EMMSAD
Workshop in Montpellier, France. We hope to write an
expanded version of that paper for a planned anthology of papers from
If you have comments or questions on the paper, or
on database schema design in general, please feel free to send me an
e-mail: firstname.lastname@example.org. I will respond to any
questions as quickly as I can.