In the recent years, object-oriented and object-relational database systems gained commercial significance. This is due to their powerful capabilities in semantic modeling, allowing for descriptive and rather "natural" representations of real life facts. In most of the systems, however, this modeling capability is restricted to the design phase. As soon as the database is completely set up, design changes are very cumbersome, if not impossible. Such changes are, however, necessary in order to adapt the databases to different applications or to keep pace with ongoing development. In the field of relational databases, an additional adaptive layer is realized in the form of the external schema or view concept. In the object-oriented area, however, such a concept has not yet been established. If it is provided at all, it offers only minimum capabilities. In literature, there is some work on the nature and functionality of object-oriented views. None of these approaches have found their way into commercial products. In fact, many approaches have very high demands on the underlying systems, which cannot be accomplished by any existing commercial system. This work examines and classifies the different approaches for object- oriented views and compares them with the abilities offered by existing database implementations. It dwells especially on the object-oriented database standard ODMG, which underlies the restrictions of a compiled and statically typed environment, just like the programming language C++. On the basis of this analysis a view concept is developed, which offers maximum functionality for distribution and integration of object bases and can nevertheless be realized on a standard-compliant C++ system. An implementation was accomplished in the context of the IRO-DB project and is documented in this work. Due to the little demands on the underlying system, this approach can be transferred to other systems as well, which is shown exemplarily for CORBA. | English |