INDEPENDENTLY UPDATED VIEWS

Independently updated views (IUVs) are updated versions of a relationa l view. Unlike traditional views, updating an IUV changes only that IU V and not the underlying database. Updates to an IUV are stored in a d ifferential table (DT) and incorporated when the IUV is accessed. Ther e is no restriction on the type of query used to define the IUV. Appli cations of IUVs include alternative scenarios for decision support, lo cal updating of distributed snapshots, and CAD versioning. IUVs allow versions to be defined on dynamic databases. This may cause inconsiste ncies because both the IUV and the base tables on which the IUV is def ined can be updated independently of one another. We introduce ''overl apping updates'' that cause such inconsistencies and present algorithm s for their detection and resolution using timestamps. These algorithm s do not impose additional overhead on the database and require only t he modification of the DT. A performance analysis indicates that the c ost of querying IUVs is reasonable for a wide range of queries, views, and access strategies. We also consider a materialized implementation of IUVs. Materialized IUVs do not require a DT, relying instead on ad ditional system columns for operation codes and timestamps. Results su ggest a mix of implementation strategies for different types of IUV ap plications, depending on the frequency and nature of queries on the IU Vs, the IUV sizes, and how actively the base tables are updated.