Several other authors, [Kim 88] and [Dittrich 1986] argue that an OODBS is a DBMS with an underlying object-oriented data model. If one takes the notion of a data model in a broad sense that especially includes the additional aspects going beyond record-orientation, this view is certainly in accordance with our approach. [Dittrich 1986] and [Dittrich 1988] introduce a classification of object-oriented data models (and, consequently, of OODBS): if it supports complex objects, a model is called structurally object-oriented; if extensibility is provided, is it called behaviorally object-oriented ; a fully object-oriented model has to offer both features. This definition also requires persistence, disk management, concurrency, and recovery; it at least implicitly assumes most of the other features (where applicable, according to the various classes); in total, it is thus slightly more liberal than our approach. However, as most current systems and prototypes do not fulfill all requirements mandated by our definition anyway, this classification provides a useful framework to compare both existing and ongoing work.
We have proposed a collection of defining characteristics for an object-oriented database system. To the best of our knowledge, the golden rules presented in this paper are currently the most detailed definition of an object-oriented database system. The choice of the characteristics and our interpretation of them devolves from out experience in specifying and implementing the current round of systems. Further experience with the design, implementation, and formalization of object-oriented databases will undoubtedly modify and refine our stance (in other words, don't be surprised if you hear one of the authors lambasting the current definition in the future). Our goal is only to put forth a concrete proposal to be debated, critiqued and analyzed by the scientific community. Thus, our last rule is:
Thou shalt question the golden rules