Clock: synchronizing internal relational storage with external XML documents

In many business settings, a relational database system (RDBMS) will serve as the storage manager for data from XML documents. In such a system, once the XML data is dissembled and loaded into the storage system, XML queries posed against the (virtual) XML documents are processed by translating them into SQL queries against the relational storage. However, for applications which frequently update their XML documents, we cannot afford to reload a complete, possibly large, document for each update, instead we must be able to incrementally propagate document updates to the stored XML data. We address the issue of correctly reflecting updates of external XML documents into the loaded XML data in a relational database system. We describe Clock, a framework for synchronizing the relational storage with updated XML documents by exploiting a metadata-driven technology. First, we propose a set of (DTD preserving) update primitives for XML documents. Second, based on the mapping between XML and the relational model, we describe the propagation of those update primitives. Validation of the updates ensures they will not violate the constraints specified by the DTD. We have implemented a working prototype of the Clock system using the IBM's XML4J parser JDBC 2 and Oracle 81. We report on preliminary experiments conducted using this prototype to analyze our algorithms in a document update setting.