Abstract

A normalization theory for XML to produce good relational database designs is presented. In the normalization approach, an already designed relational database is given as input, together with some semantic information provided by a user in the form of relationships between different parts of the database, called data dependencies. A normal form, defined as a syntactic condition on data dependencies, specifies a property that a well-designed database must satisfy. XML has emerged as the standard data model for sorting and interchanging data on the Web. It is shown that XML documents may contain redundant information, which could be related to the hierarchal structure of these documents. A functional dependencies language for XML is introduced, which is used to define XNF, a normal form for XML documents. As the new data formats such as XML are becoming critically important, classical database theory problems have to be revisited in the new context.