James Clark announced an update of the TREX web site with refinements, and has created a complete implementation of XHTML modularization in TREX.
The only change made to the TREX language is a refinement to the way it handles merging included grammars.
Following hard on the heels of Murata Makoto's XHTML endeavors in RELAX, Clark has hand-crafted an implementation of the W3C's XHTML Modularization Candidate Recommendation in TREX.
XHTML Modularization decomposes XHTML 1.0 into abstract modules providing specific types of functionality. Clark writes of his implementation:
This has been done by hand not automatically, so there are probably
bugs, but it takes full advantage of TREX. Notably, with TREX there's
no need to create a model module specific to the combination of modules
you are using; you just have to include the modules you need. I was
also pleasantly surprised to discover that it was possible to implement
exceptions in TREX (used in HTML 4 to, for example, prevent nested
links), even though there's no specific exceptions feature in TREX.
Clark has added an explanation of the exception mechanism into his TREX tutorial.