I gave up trying to get the HTML to look nice for the moment and went back to testing my purified datatypes, which threw up the following oddity. The attribute @prefix on <elementSpec> is currently defined (in RNG) as follows <rng:choice> <rng:value/> <rng:ref name="data.xmlName"/> </rng:choice> (or in compact form: attribute prefix { "" |data.xmlName <http://www.tei-c.org/release/doc/tei-p5-doc/en/html/ref-data.xmlName.html> }? ) I*think* this is so that you can distinguish (say) <elementSpec ident="foo" prefix=""/> from <elementSpec ident="foo" > but my question is WHY WOULD YOU WANT TO DO THAT? If no-one can come up with a persuasive case where this distinction might be necessary, I'd rather change this to <dataRef key="teidata.xmlName"/> than go to the trouble of defining a new <dataSpec> for this case. If I make that simplification, prefix="" would then be illegal; but wouldn't it be just as effective to say that if no prefix is supplied the implication is that the prefix is null?