When I first became involved in data modeling in the mid-1970s, I was taught a set of diagramming conventions, the rules of normalization, and a few principles of good design. It did not take me long to discover that my education had covered only the easy part. The real challenge, as any experienced modeler knows, lies in understanding business requirements and choosing an appropriate set of concepts and structures to support them. The traditional advice to “ask which things the enterprise needs to keep information about and how they are related” is a gross over-simplification of the often very difficult process of identifying entities and relationships.
Research in the last few years has supported what practitioners have known for a long time: rather than modeling from first principles, experienced data modelers re-use and adapt models and parts of models from their previous work. In fact, their “experience” may well reside more in their personal library of models–typically remembered rather than documented–than in greater facility with the basic techniques. The use of pre-existing templates also changes the nature of the dialog between the business experts and modelers: modelers will seek to discover which model or models from their repertoire may be appropriate to the situation, then to check the detail of those models. This is a far more proactive role for modelers than that traditionally described, and recognizes that both parties can contribute ideas and content to the final model.
Of course, it takes time and exposure to a wide variety of business requirements for an individual to build up anything approaching a comprehensive library of models. Only specialist data modelers are likely to have this opportunity, and the reality is that much data modeling is performed by non-specialists.
The obvious step forward from this rather haphazard individual approach is for experienced modelers to develop and publish models for the most commonly encountered business requirements, so that solutions can be shared, reviewed and improved. Almost every commercial enterprise needs to keep data about customers, about staff, about sales. And almost every data modeler has spent time wrestling with these common–but by no means simple–situations, painfully aware that he or she is re-inventing the wheel, but without any confidence that any particular modeler has done a better job.
Such additions to data modeling's “body of knowledge” have been a long time coming. Books, papers, and educational material have continued to focus on the foundations of data modeling: modeling paradigms, diagramming conventions, and normalization. These are important topics, to be sure, but the absence of more developed material lends credence to the argument that data modeling does not deserve the status of a fully-fledged discipline.
Perhaps the reason for the gap in the literature is that the individuals best placed to recognize common situations and to develop models for them are data modeling practitioners–more particularly consultants who have had the opportunity to see a range of different business requirements. The models that they have developed over the years are a valuable professional resource, more profitably deployed on consulting assignments than as material for general publication. It also takes some courage to present one's own solutions for scrutiny by peers, all of whom will turn naturally to the problems for which they have personally developed the most elegant solutions!
I am therefore delighted that Len Silverston has chosen to publish a second and substantially expanded edition of The Data Modeling Resource Book. The first edition was essential reading for anyone charged with developing data models for business information systems, and was particularly notable for including contributions by specialists in particular data modeling domains. The second edition retains this feature, covers new business areas, and updates the original material. Len's willingness to continue to improve the material gives me hope that the core models will acquire a deserved status as standard starting points.
The second edition of The Data Modeling Resource Book is an excellent answer to the question “what is the second data modeling book I should purchase, once I've learned the basics?”–and every practitioner of data modeling should own at least two books on the subject!
Graeme Simsion
1 January 2001