this paper. The proposal extends the key speci cation of XML Data by allowing one to specify keys in terms of XPath [24] expressions. There are a number of technical problems in connection with XPath. XPath is a relatively complex language in which one can not only move down the document tree, but also sideways or upwards, not to mention that predicates and functions can be embedded as well. The problem with XPath is that questions about equivalence or inclusion of XPath expressions are, as far as the authors are aware, unresolved; and these issues are importantifwewant to reason about keys as wedo in relational databases. Yet until we know how to determine the equivalence of XPath expressions, there is no general method of saying whether two such speci cations are equivalent. Another technical issue is value equality. XML Schema restricts equality to text, but the authors have encountered cases in whichkeys are not so restricted. See Section 7.1 for a more detailed discussion. However, the main reason for writing this paper is that none of the existing key proposals address the issue of hierarchical keys, which appear to be ubiquitous in hierarchically structured databases, especially in scienti c data formats. A top-level key may be used to identify components of a document, and within each component a secondary key is used to identify sub-components, and so on. Moreover, the authors believe that the use of keys for citing parts of a document is suciently important that it is appropriate to consider key speci cation independently of other proposals for constraining the structure of XML documents

As XML becomes a standard for data representation on the internet, there is a growing interest in storing XML using relational database technology. To date,

Much work on XML data was around storage and querying and did not consider constraints of XML, especially keys. Since constraints have been proposed in many papers for XML, much research work on constraints has been being done. In this paper, we consider an important class of constraints, XML keys, and try to find the relationship between XML keys and relation keys. Given XML data whose semantics are represented in XML keys, we use a simple transformation language to express a transformation from the XML data to a relational database. Then we provide a formal definition of a relation tree that represents a well-formed rule on a relation written in the transformation language. After all rules that make up the transformation have been written, we show the way how XML data is transformed into relations. Finally, we present an algorithm for deriving relation keys from XML keys. These keys should be specified on the transformed relations or their compatibility with the predefined schema should be verified.