Data Modeling using XML Schemas

Ph.D. Candidate

Murali Mani
UCLA

April 30, 2003
11 a.m. - 12 noon
Fuller Labs 320

Abstract

A revolution occurred in 1998, when XML 1.0 was published by W3C. XML 1.0 allows user-defined tags to represent semantics. This was a simple, yet powerful, extension to HTML, which combined with the standard encoding schemes for data exchange, has made XML applicable to wide variety of data exchange applications. Today, XML has established itself as the lingua franca for information exchange over the web.

In addition to its universal impact in information exchange, XML created a huge impact in database applications also. From 1970, database technology has been based on the first normal form proposed by Codd. There were several attempts to move away from first normal form, such as nested relational model, from which we have learnt significantly, but which have not been very successful in practice. Some of the features of object-oriented model have been incorporated into traditional database technology, but object-oriented models as such have not been very successful in practice. The advent of XML provided the break-through several database researchers and practitioners have been looking forward to - provide a non-first normal form logical model, which is easier for the user for modeling database applications, without losing the independence between logical and physical models.

In this talk, we will describe how relational databases can be published as XML. We will first examine how structures and constraints are specified in the XML model, and different options for specifying them. We will then describe how entities and relationships can be represented in XML. After getting a clear understanding of these aspects, we will describe three algorithms for publishing relational databases as XML. The main features of the algorithms are they maintain semantic constraints, they require minimal user interaction, and the resulting data model does not introduce new redundancies, rather they remove some redundancies present in the original relational model. We will present some experimental observations, and some of our experiences in XML data modeling requirements.

Host

Prof. Micha Hofri

Refreshments will be served.

Maintained by webmaster@wpi.edu
Last modified: Sep 27, 2006, 16:05 EDT