Markus Tresch

Updatable Views in Object-Oriented Databases

Object-oriented database systems (ooDBMSs) are supposed to offer at least the functionality available in commercial relational DBMSs of today. One important consequence of this is that they have to provide a separation of the global (conceptual) database schema from the external schema (“subschema”) of a particular task. Views are a mechanism to realize this data independence. In addition, they also support multiple levels of detail, security and authorization, and interoperability in a heterogeneous environment. In a relational DBMS, views are defined by queries. However, they can not be freely updated. We describe concepts of an object model and query language that are necessary for object view definitions. We show that updating object views is much more feasible than in the case of relational views. The key property of a query language leading to this result is object preserving operator semantics. That is, in contrast to many previous object algebras, query results are sets of existing objects instead of data tuples or new objects. Consequently, we have to solve the classification problem: where to include the view in the type and class lattices.

Schema transformation without database reorganization

We argue for avoiding database reorganizations due to schema modification in object-oriented systems, since these are expensive operations and they conflict with reusing existing software components. We show that data independence, which is a neglected concept in object databases, helps to avoid reorganizations in case of capacity preserving and reducing schema transformations. We informally present a couple of examples to illustrate the idea of a schema transformation methodology that avoids database reorganization.

Meta Object Management and its Application to Database Evolution

In this paper, we address the problem of supporting more flexibility on the schema of object-oriented databases. We describe a general framework based on an object-oriented data model, where three levels of objects are distinguished: data objects, schema objects, and metaschema objects. We discuss the prerequisites for applying the query and update operations of an object algebra uniformly on all three levels. As a sample application of the framework, we focus on database evolution, that is, realizing incremental changes to the database schema and their propagation to data instances. We show, how each schema update of a given taxonomy is realized by direct updating of schema objects, and how this approach can be used to build a complete tool for database evolution.

The COCOON object model

The COCOON model was intended to extend the concepts of relational database management systems DBMSs beyond nested relational to object oriented ones Key characteristics of COCOON and its database language COOL are generic set oriented query and update operators similar to relational algebra and SQL updates respectively object preserving seman tics of query operators which allows for the de nition of updatable views update operations that keep model inherent integrity constraints consis tent a separation of the two aspects of programming language classes type vs collection predicative description of collections similar to de ned concepts in KL One like knowledge representation languages automatic classi cation of objects and views positioning in the class hierarchy This report gives a comprehensive introduction to the COCOON model and its language COOL as well as a formal de nition Our formalization uses de notational semantics a popular technique in programming languages We found that standard set theoretic formalizations of data and object models were not equally well suited to express update semantics This however is essential for object oriented as opposed to value based languages This can also be taken as an indication for the convergence of database languages to generell programming languages Along these same lines we emphasized static type checking of COOL Second revised edition January

Towards a platform for distributed application development

This paper describes the architecture of a generic platform for building distributed systems over stand alone applications. The proposed platform integrates ideas and technology from areas such as distributed and parallel databases, transaction processing systems, and workflow management. The main contribution of this research effort is to propose a “kernel” system providing the “essentials” for distributed processing and to show the important role database technology may play in supporting such functionality. These include a powerful process management environment, created as a generalization of workflow ideas and incorporating transactional notions such as spheres of isolation, atomicity, and persistence and a transactional engine enforcing these “quality guarantees” based on the nested and multi-level models. It also includes a tool-kit providing externalized database functionality enabling physical database design over heterogeneous data repositories. The potential of the proposed platform is demonstrated by several concrete applications currently being developed.

Schema Transformation Processors for Federated Objectbases

In contrast to three Schema levels in centralized object-bases, a reference architecture for federated objectbase Systems proposes five levels of Schemata. This paper investi-gates the fundamental mechanisms to be provided by an object model to realize the processors transforming be-tween these levels, namely Schema extension, Schema fil-iering, and Schema composition. It is shown, how compositum and extension are used for stepwise bottom-up integration of existing objectbases into a federation; and how extension and filtering support authorization on dif-ferent levels in a federation. A powerful View definition mechanism and the possibility to define subschemata (i.e., parts of a Schema) are the key mechanisms used in these processes.

A classification of multi-database languages

This paper defines a formal classification of multi-database languages into five levels of database integration with increasing degree of global control and decreasing degree of local autonomy. First, the fundamental interoperability mechanisms are identified for each of these levels. Their consequences on local autonomy as well as implementation drawbacks are discussed. Second, various multi-database languages are classified into these categories. In addition to COOL*, other proposals are analyzed, including SQL*Net, Multibase, Superviews, VODAK, Pegasus, and O*SQL.<<ETX>>

Evolution towards, in, and beyond Object Databases

There is a manifold of meanings we could associate with the term “evolution” in the database arena. This paper tries to categorize some of these into a unique framework, showing similarities and differences. Among the topics touched upon are: extending traditional data models to become “object-oriented”, migrating existing data to (not necessarily OO) databases, schema extension and modification in a populated database, integration of federated systems, and the use of “external services” to enrich DBMS functionalities. The following are presented in more detail: first, we describe the necessity of object evolution over time; second, we discuss schema evolution; and third, we present evolutionary database interoperability by identifying different coupling levels. A few basic mechanisms, such as views (derived information) and a uniform treatment of data and meta data, and type and/or class hierarchies, allow for a formal description of (most of) the relevant problems. Beyond presenting our own approach, we try to provide a platform to solicit further discussion.

An extensible classifier for semi-structured documents

In this paper, we present a vector space classifier for determining the type of semi-structured documents. Our goal was to design a high-performance classifier in terms of accuracy (recall and precision), speed, and flexibility. The ability to dynamically extend a classifier with userspecific classes is crucial for many applications. Unfortunately, the training data of existing classes is often not available, such that the extended classifier is imprecise a+ a result. We focus on this issue. First, we evaluate how to create class abstracts that can be used as training data replacement. Second, we introduce relevance feedback learning strategies to overcoming the remaining classifier flaw.