Classifying and Resolving Unique Constraint Violations during Synchronization Processes

In mobile environments data from a client database often has to be integrated into a server database during a synchronization process. Thereby unique constraint violations are possible if the client has older data or only part of the data available offline. When inserting data into the server database in an online scenario, a unique constraint violation can be recognized and corrected immediately. However, when being offline a possible later violation might go unrecognized because not all data is available to check against. Later, when synchronizing or merging an offline database with the server database, a violation will occur and the synchronization process will have to handle it. Many synchronization and optimistic data replication approaches treat a unique constraint violation as a conflict. The advantage of this treatment is that the same logic can be used for resolving conflicts and unique constraint violations. However, sometimes conflict resolution can only resolve a violation by overwriting an existing version, or it is not able to resolve the violation at all. By implementing a discrete unique constraint violation resolution, the available information is utilized and overwriting data can be avoided. In most cases the violation can even be resolved automatically. This paper identifies two classes of unique constraint violations and provides a resolution approach for each class. The classification is based on the experiences collected in a project for a social insurance company in which such violations occurred during the synchronization process. These violations were examined in detail and are presented in this paper.