Isn't schematron powerful enough in its niche? Possibility to validate dtd-like expressed content model adds no real value to schematron, whose power is in ability to express constraints that are unexpressible in dtd or schema. Take notorious cooccurence constraints as example.