I added full-text search to my Python/XML and 4Suite "Akara" sites (my document collation projects). As I predicted, it took less than an hour.
Dave Carlson pointed me to his article: Modeling XML Applications. I like his thinking in general, though I do have some specific qualms. For one thing, I don't think XML schemata are really solid information modeling bases for the model in the MDA. They're text tree modeling systems, with primitive type inferencing, and we need something more powerful. I, of course, advocate RDF, but even Topic Maps are more suitable. I notice that Dave credits WXS for being the message content model system built into SOAP, but it's actually not WXS but a specialized SOAP serialization, which has some superficial similarity to WXS but is fundamentally quite different.