Normalization / Marc Rettigs poster
Normalization once and for all!

One of the "data modelling classics" is a famous poster issued by Database Programming & Design in 1989 about normalisation. The author is Marc Rettig - the then technical editor. Reused here with the kind permission of Marc, who is now with fit associates. The poster is a very good presentation of normalisation taken to the fifth normal form. © Marc Rettig, 2006

Data Modelling



The Return of the Modellers!


Data Mod
elling is back in focus. After a number of years, where Data Modelling was simply considered a well understood best practise, its importance is reemphasized because of factors such as:

  • Serviceorientation and other integration projects
  • Data Governance
  • Master (Reference) Data Management
  • Enterprise 2.0 and Semantic Integration
  • Inclusion of unstructured data / “enterprise search”

So, let us take a look at the data modelling universe:

I am not going to teach you data modelling in this place - there are lots of other places, where you can find that. But there are a few issues, I would like to point your attention to.
  1. Business people describe their business concepts
  2. Information Architects facilitate that process
  3. Information Architects design Logical Data Models based on their understanding of the business
  4. Data Modellers specify Database Datamodels for the DBMS of choice
  5. Developers implement on top of the above.
This the flow. Many it people forget the first two bullets or get them messed up with the third one.

Business Concept Modelling


Seems to me that what was once known as Conceptual Modelling has ceased to exist. With no good reason. There is a lot for information modellers to do on the business level. The models at this level are imprecise (fuzzy) and use the business language. See my discussion about Business Concept Modelling here.

Logical Datamodelling


This level is a DESIGN level! Do not forget that. It is heavily influenced by very rich and very rigid formalisms such as UML, 3 NF, RDF/OWL etc. Lots of design decisions are being made and are built into the models.

CRUDS - the business lifecycle


Another thing, which is almost completely forgotten today, is the life cycle analysis of business object types. Remember it? It is all the business rules concerning:
  • Create
  • Read
  • Update
  • Delete
  • Search
These issues are fighting their way back into information management by way of piggybacking Master Data Management. A good thing, for sure.

Multidimensional denormalization


Today, many data modellers have learned to denormalize with a smile (the words of Ralph Kimball). Building Star Schema models (and OLAP cubes) is sort of turning the database “inside out”. And it has been a great succes! Not many steadfast believers in the 3rd normal form back in the eighties would have believed that! But it was a good thing that we learned how to do it. (This publisher has personally trained at least 600 people in multidimensional modelling - and it has been great fun).

Normalization


This technique hasn’t changed a whole lot. I include it here to draw your attention to the famous normalization poster published in 1989; see the sidebar to the right!

Cognitive Science


Finally, things are happening again in the area of “artificial intelligence”. That term is almost complete dead, and is replaced by Cognitive Science. Se my discussion of it here.