Virtual XML and rewriting XQueries for heterogenous data sources
Kristoffer H. Rose, IBM Thomas J. Watson Research Center

Virtual XML is about using XML as a universal data model for all data. I'll give an overview of how this works and why it is feasible in practice, specifically how adaptive wrappers can be written such that data sources with different access profiles as well as structure can be processed as if they were complete in-memory XML documents, even when the virtual documents are very large (such as entire file systems and the web as single documents). I will show examples, and justify why the XML approach can actually work as a universal data model.

In such large scale situations it is a challenge to make query optimization work: we essentially need to apply database optimization as well as traditional functional optimization while respecting data access profiles. To this end we have developed a generic higher-order rewrite engine powerful enough to express both kinds of rewrites, as well as permit incorporation of analyses and semantics-based static reduction. I will show how the rewrite engine works and give examples of how it is used to implement combined optimizations.

References:

• http://www.research.ibm.com/virtualxml
• http://crsx.sourceforge.net