Well then, I think you should come up with a good proposal on how to
derive an SGML declaration from the charset parameter, and a good
proposal for handling numeric charater references in the face of blind
encoding transformations.
Can you give me some examples of characters sets people think are
richer than ISO 10646?