抄録
This paper discusses the usability of convertibility, a principle for data quality used by the Xplain-DBMS. Convertibility (uniqueness) of type definitions is a helpful criterion for database design, whereas convertibility of instances is a criterion for the uniqueness of instances (records). However, in many situations with or without generalization/specialization, convertibility appears to be an insufficient criterion for correctness of instances, which is illustrated by many examples. In order to be able to specify more rigorous rules for correctness of instances we propose to use new concepts such as 'identifying property'. These new concepts also facilitate the transformation of relational databases into Xplain databases.
