In modern universal database management systems (DBMSs) user-defined data types along with extensible indexing structures try to bridge the gap between standard data-independent DBMS implementations and the requirement of specialized access methods for efficient domain-specific data retrieval, maintenance, and storage. However, these approaches often suffer from restricting the degree of freedom in the implementation and limiting the availability of crucial database features. Due to their design concepts, these extensible indexing frameworks are not intended to be suitable for rapid development and evaluation of research prototypes, as they lack essential generalization, completeness, and depth of their integration into the host DBMS. We discuss the advantages and drawbacks of available extensible indexing techniques and present several methods that can be easily combined into a powerful and flexible framework for storing, indexing, and manipulating domain specific data from any data source. We demonstrate that this framework comprises all properties truly extensible indexing should have. A prototype implementation of this framework was integrated into the relational DBMS Transbase®.
展开▼
