XML Databases and Beyond - Plenty of Architectural Challenges Ahead

A key observation is that the invariants in database management determine the mapping steps of the supporting architecture. Referring to the multi-layered architecture of record-oriented database management systems (DBMSs), we sketch the advances made during the past decades. Then, we explore the ways how this proven architecture can be used to implement XML DBMSs (XDBMSs). Major changes and adaptations are needed in most of the layers to support fine-grained XML document processing (XDP). The use of DeweyIDs opens a new paradigm for the management of XML document trees: While preventing node relabeling, even in case arbitrary large subtrees are inserted into an XML document, DeweyIDs offer great benefits for efficient navigation in the document trees, for declarative query processing, and for fine-grained locking thereby avoiding access to external storage as far as possible. The proposed architecture also captures horizontal and vertical distribution of XML processing. Nevertheless, new architectural models are needed beyond record-oriented data types.