Keesee: A Behavioral Object-Oriented Database Framework for Software Engineering

Abstract

In this paper, we discuss a research project, KeeSee (KEETM2 Software Engineering Environment), which is concerned with the support of software engineering environments (SEE’s) by database systems. The general topic of database support for software engineering has recently received significant attention from the database community. In these projects it has been frequently noted that software engineering (and engineering in general) places a new set of constraints on a DBMS. A common theme in database research is that existing data models do not suffice to support the following characteristics of SEE’s: long transactions, hierarchically structured objects, complex derived data, and version management. Current efforts to address the bottleneck in existing data models can be classified into two categories: object oriented systems and relational extensions. The prototype KeeSee software is implemented in KEE, an object-oriented environment for constructing expert systems. KEE provides facilities for inheritance, hypothetical databases, rules, demons, queries, message passing, and truth maintenance. The KeeSee project makes particular use of the multiple inheritance in KEE and message passing in defining database functionality to support a SEE. Essentially, one defines the necessary database functionality through the multiple inheritance capabilities in KeeSee. Then, software tools are integrated by defining additional methods for object types. In this manner, the advantages of object-oriented programming are used to both define the database functionality and also to incorporate software tools. As a result, the KeeSee system has turned out to be easy to use and modular for implementers and users of the SEE. There are two main results from this study. First, the breadth of functionality in KEE is an essential component for supporting future SEE’s. Second, method inheritance and object specialization provide a powerful mechanism for defining the SEE as well as integrating tools. Combining these results with current trends in database support for software engineering suggests that an efficient implementation of a system like KEE, combined with database technology is an important component of the ultimate platform for software engineering.