|
|
|
This paper is an excellent survey of temporal extensions to the well-known Entity-Relationship (ER) model [2]. The ER model is probably the most widespread conceptual modeling notation in both research and industry. However, the temporal aspects of complex domains are difficult to capture with the basic ER model. For example, entities may have a lifespan, relationships may change over time, attributes of both entities and relationships may have a time validity, etc. Also, these intervals may be expressed in different time granularities, and other dimensions of time may be relevant for certain applications (e.g., transaction time). Very often these temporal aspects are ignored, or only partially considered during conceptual modeling, and then hidden in the implementation of the specific application. Clearly, without proper modeling, the resulting application may be difficult to manage and may hide unintended behaviors.
If temporal aspects are to be modeled, it is still under debate if this should be done while modeling the basic entities and relationships, or only after a complete non-temporal ER schema has been constructed. For example, [3] takes this second choice, adding in a second phase ``annotations'' to each ER component with a description of its temporal semantics. (In [3] these annotations are external to the ER model.)
Each proposal considered by the survey illustrates an extension to the ER syntax and/or semantics in order to nicely capture within the extended ER model some of the temporal aspects. In some cases the extended ER can be later translated in a standard one, enabling automatic mapping to a relational schema. Some of the extended ER models actually enable the approach taken in [3], while others imply temporal modeling from the beginning. The different proposals are compared against a number of desirable properties which the authors have identified. There is no winner approach satisfying all criteria, but expressiveness, advantages, and problems are carefully evaluated. What probably distinguishes this survey most from related studies is that the emphasis is kept on structural aspects and conceptual modeling capabilities, while the underlying data models and query languages play a minor role.
I believe the paper is a good reference for all people working in temporal databases and conceptual modeling in general.
Copyright © 2000 by the author(s). Review published with permission.