|I'm using Dan's framework of talking about the SGML document
|character set as a "coded character set" : a mapping from integers
|to characters (unrelated to the MIME character encoding).
|Thus, every character in the document character set, has
|one or more code positions in this mapping, regardless of
|if it actually appears in a numeric character reference.
|Those code positions (or whatever we want to call them) determine
|the value of numeric references to characters, but their
|existence is not contingent on the actual use of numeric references.
|For document character sets whose code domain and character range
|_are_ a subset of ISO 10646 it is my intention that my
|proposal be equivalant to the last proposal: the document
|character set must be ISO 10646 or the restriction of it to
|an appropriate subset of characters.
|The special cases arise only if one wants to use characters
|not in ISO 10646 or code positions unused in ISO 10646: this
|is allowed but folks are warned to be careful of the consequences.
OK, then I simply misunderstood (as I was hoping). You're just saying
that the document character set is ISO 10646. The three rules you used
to state that are simply a bit verbose. The bit about using unassigned
code positions for private use is redundant, since that is part of the
definition of ISO 10646.
-- Bert Bos Alfa-informatica <email@example.com> Rijksuniversiteit Groningen <http://www.let.rug.nl/~bert/> Postbus 716, NL-9700 AS GRONINGEN