The thesis of this research is that XML is a structured representation of information, and databases are used for storage and retrieval of structured information, so it is possible to map XML documents into a database and vice versa. Done properly using mature database software, this allows efficient processing of XML data. The authors actually focus on two partial mappings in this paper. One mapping is from an XML document to four relational tables, and the other mapping converts XPath query expressions into corresponding SQL queries.
The authors implemented their design and compared the resulting databases and SQL query times to a system designed by Florescu and Kossman (FK99). Within the scope of this review, the important consideration is that the FK99 design involves 22 tables corresponding to the specific structure of the XML data used. The authors’ approach is fundamentally different, allowing their four tables to represent any XML document in storage space comparable to the more complicated designs.
Evaluating the results of the query times is difficult. There are many kinds of queries, and the query forms and search times strongly depend on the structure of the data, which are very different. While their query response times compare very favorably with FK99, they only report on eight queries, and it isn’t even evident if those are representative of the kinds of queries that might be made against this data.