The many people and organizations who came to the XML industry from the database and software development industries have always wanted better standards for modeling the native data structures they interchange in XML. Some support for this was always likely in developing the XML Schema language that was expected to supplant XML 1.0 DTDs. Yet when drafts of the W3C XML Schema language (WXS) emerged, the mechanism provided for integrating with data type support proved immediately controversial.

To simplify the matter a bit, the Post Schema Validation Infoset (PSVI) annotates the Infoset of an XML document (i.e., the abstract model of its nodes) with information about the types that became associated with each node during schema processing. The many people who came to XML from the document and text management space, and even from the Web architecture space, objected to this imposition of programming types into XML at such a fundamental level. The main fear was that this would end up causing a lot of complexity in other XML technologies even if people didn't use XML as a mere data serialization format.

The introduction of the PSVI is one of the reasons that so much interest persisted in alternate schema languages. RELAX and TREX, which merged to form RELAX NG, limited themselves to validating documents. RELAX NG allows one to assign data types from plug-in libraries to simple element content and attribute values, but it doesn't provide a model of any sort for these types. Such a model would have to be layered upon the process that does the validation. This difference has led to a palpable split in the XML community between those who can't imagine XML without built-in provisions for data typing and those who can't imagine why such things shouldn't be layered very distinctly from XML's textual core. Furthermore, one camp sees data typing as needed for XML tools, while the other worries that the added complexity will make XML impossible to process without sophisticated (and thus expensive) tools. The open source versus commercial software battle thus enters into the mix.

There is actually more to the choice between RELAX NG and WXS than data typing: RELAX NG has gained unexpected momentum largely because it's a much earlier tool for modeling common XML constructs. However, the debate between XML-as-pure-text and XML-as-serialized-data has become more prominent as the PSVI wends itself into other specifications. The emerging new battlefield is the W3C XML Query language and its effect on the next iterations of XPath and XSLT.

XPath and XSLT are arguably the most successful XML-related technologies besides XML 1.0 itself - and they're relatively simple, providing ways of addressing the nodes that make up XML documents pretty much as they are after basic parsing. Some facilities, such as the sum() function in XPath and in XSLT, do provide for superimposing operations based on data types, but these are lightweight. Users who take to WXS and thus end up using its sophisticated data-typing system, including such things as dates and floating-point numbers, wish for more tools for manipulating these in XPath and XSLT.

The XML Query working group early on recorded a requirement for support of WXS types. XQuery modeled itself as a sort of supercharged extension of XPath node access and XSLT result templates, and eventually the W3C decided to pool the effort toward XQuery 1.0, XPath 2.0, and XSLT 2.0. The result of all this was that next-generation XPath and XSLT drafts became much more complex as they included all manner of facilities for PSVI and data types.

People in the XML-as-text camp, who at first simply wrote off XQuery as a byzantine curiosity that they needn't worry about, suddenly found their beloved XPath and XSLT transformed into exactly the kinds of specifications they feared the PSVI would inspire.

The result of all this right now is a pitched battle between the text and data types camps, with XQuery held up as the model for all that is supposedly good or bad about the directions in which XML is moving. XQuery itself has been undergoing changes to try to accommodate all the competing concerns. And comments on most of XML's public mailing lists, including that of the W3C technical architecture group, intimate that a showdown may be in the offing as the drafts in question progress toward recommendation.

XML technology has brought together an impressive diversity of practitioners. Now a debate between the two prominent branches is destined to shape its future.