Making the Infoset Extensible

The XML Infoset defines the data model of XML, and it is used by a number of other specifications, such as XML Schema, XPath, DOM, and SAX. Currently, the Infoset defines a fixed number of Information Items and their Properties, and the only widely accepted extension of the Infoset are the Post Schema Validation Infoset (PSVI) contributions of XML Schema. XML Schema demonstrates that extending the Infoset can be very useful, and the PSVI contributions of XML Schema are being used by XPath 2.0 to access type information in a document's Infoset. In this paper, we present an approach to making the Infoset generically extensible by using the well-known Namespace mechanism. Using Namespaces, it is possible to define sets of additional Information Items and Properties which are extending the core Infoset (or other Infoset extensions, defining a possibly multi-level hierarchy of Infoset extensions). Basically, a Namespace for an Infoset extension contains a number of Information Items, which may have any number of Properties. It is also possible to define an Infoset extension containing only Properties, extending the Information Items of other Infosets. Further elaborating on this method, many of the XML technologies currently using the Infoset could be extended to support the Infoset extensions by importing Infoset extension using the extension's Namespace name. To illustrate these concepts, we give an example by defining the XML Linking Language (XLink), the XML vocabulary for hyperlinking information, in terms of Infoset extensions. We show how the proposed ways of supporting Infoset extensions in XML technologies such as XPath, DOM, and CSS could pave the path to a better support (and hopefully faster adoption) of XLink than we see today. XLink serves as one example, but the proposed extensions and techniques are not limited to this particular technology. The content of this paper is work in progress, contributing to the ongoing debate on how to deal with different XML vocabularies and their usage in other XML technologies. We believe that making the Infoset extensible would provide a robust and flexible way of making the data model of XML-based data more versatile, and creating an accepted way of making the data available through standard interfaces such as DOM and XPath.