In the announcement, Clark indicates that the main change to the implementation is that it now supports datatyping using a "real implementation" of the W3C's XML Schema Part 2: Datatypes Candidate Recommendation. (Note that the TREX language itself does not have any system of datatypes built in, but is instead designed to "partner" with a datatyping vocabulary such as the W3C's XML Schema Part 2: Datatypes.) The implementation does not support every datatype in the W3C Candidate Recommendation; Clark provides a document that details which datatypes are supported and which are not.

The hallmarks of TREX, which Clark introduced publicly only one month ago, are its simplicity and flexibility. It attempts only to specify patterns for defining the tree structure and content of specific classes of XML documents, without aiming to assist in interpretation or processing of the documents. The TREX site includes both a tutorial that provides extensive examples of the language, and a formal specification that details its syntax and semantics.

TREX figured prominently in a recent xml-dev discussion thread (which Schematron creator Rick Jelliffe started with the subject line "Are we losing out because of grammars?") discussing the merits of, and differences between, various XML schema languages (including Murata's RELAX, Jelliffe's Schematron, and the W3C's XML Schema language in addition to TREX). Key parts of the discussion are very nicely summarized and elucidated by Leigh Dodds in his February 7th XML-Deviant column.

Clark also contributed another recent message to xml-dev in which he provides details about "areas where TREX allows things that are difficult or impossible with the W3C's schema language".

Related stories: