When a database is shared by many users, updates to the database schema are almost always prohibited because there is a risk of making existing application programs obsolete when they run against the modified schema. The paper addresses the problem by integrating schema evolution with view facilities. When new requirements necessitate schema updates for a particular user, then the user specifies schema changes to his personal view, rather than to the shared base schema. Our view schema evolution approach then computes a new view schema that reflects the semantics of the desired schema change, and replaces the old view with the new one. We show that our system provides the means for schema change without affecting other views (and thus without affecting existing application programs). The persistent data is shared by different views of the schema, i.e., both old as well as newly developed applications can continue to interoperate. The paper describes a solution approach of realizing the evolution mechanism as a working system, which as its key feature requires the underlying object oriented view system to support capacity augmenting views. We present algorithms that implement the complete set of typical schema evolution operations as view definitions. Lastly, we describe the transparent schema evolution system (TSE) that we have built on top of GemStone, including our solution for supporting capacity augmenting view mechanisms.

No abstract available.

Simulatingschema evoiution using views offers many advantages over the direct modification of the schema.However, this view approach towards transparent schema evolution has not yet been implemented.One reason for this may be that views can 't simulate capacity-augmenting schema changes due to the inherent limitation of view mechanisms that do not augment the underlying schema.One potential solution for this problem is to develop capac@- augmenhngview support.However, existing 00 view mechanisms do currently not support this capacityaugmenting feature.This paper thus proposes an alternative, more practical, solution to this problem.This solution is neither confined to object-preserving schema changes nor requires capacdy-augmentmg views.Note that our proposed solution is the first to aJso support hard changes such as converting values into an object, splitting two classes vertically, etc., in a transparent fashion.We demonstrate the feasibility of our solution by presenting general algorithms for hard schema changes as well as primitive ones.

When a database is shared by many users, updates to the database schema are almost always prohibited because there is a risk of making existing application programs obsolete when they run against the modified schema. This paper addresses the problem by integrating schema evolution with view facilities. When new requirements necessitate schema updates for a particular user, the user specifies schema changes to the personal view rather than to the shared base schema. Our view evolution approach then computes a new view schema that reflects the semantics of the desired schema change, and replaces the old view with the new one. We present algorithms that implement the set of schema evolution operations typically supported by OODB systems as view definitions. This approach provides the means for schema change without affecting other views (and thus without affecting existing application programs). The persistent data is shared by different views of the schema, i.e., both old as well as newly developed applications can continue to interoperate. In this paper, we present examples that demonstrate our approach.>