>I'm not sure, but Rick Jelliffe has done some work on entity sets. I
think he had one very good suggestion too: rather than use numeric
character references, why not use/adopt the practise of referring to
Unicode character via a U+hexcode; entity reference. It would be
trivial to generate a full entity set for anyone wishing it. This is
far ore obvious than numeric character references, and somewhat more
robust in faulty parsers as well.
I don't see how this is useful. You still have to have tens of
thousands of entities, and now instead of mnemonic names they
have meaningless-to-most-humans names. The numeric charrefs
are no less meaningless and we get them for free with the
change in the SGML decl. Nobody has to lift a finger to construct
an entity set.
I agree that mnemonic entities are useful, and I'd like to see
the 8879 sets supported, but let's not deprecate the numeric
charrefs, which i18n-conformant HTML browsers will have to support.
Terry Allen ([log in to unmask]), Online Books Editor, Songline Studios
affiliated with O'Reilly & Associates, Inc.
A Davenport Group sponsor. See http://www.ora.com/davenport/README.html
None genuine without .signature