On 21/11/2012 21:48, Sebastian Rahtz wrote:

The Text Encoding Initiative (http://www.tei-c.org/) manages this sort of thing by defining a whole new higher-level XML language of its own for describing the various areas it covers (dictionaries, plays, manuscript descriptions, transcriptions of spoken speech etc). This language includes mechanisms for subsetting, group, classes of objects etc. All the processing is done using XSLT, to generate DTD or RELAX NG schemas (from which XSD can be generated) or extracted ISO Schematron or, documentation.

I don't particularly suggest you adopt the TEI ODD language for your own work (fun though it would be), but just to point out that there are precedents for large-scale generation of schemas

Why don't you recommend TEI ODD? - Olumide