hallucinatory blueberry decoupling

["Simon St.Laurent" <simonstl@simonstl.com>]

The Blueberry issues seem to have exposed a fault line in XML, where the
XML plate and the Unicode plate rub up against each other and sometimes
shift violently.  Blueberry seems like an effort to address the movement
of these plate, but I'm not sure that the proposals made thus far really
address the underlying issues of the two separate and separately-managed
specifications.

I'm wondering if binding the character class lists into XML 1.0 was a
mistake.  I'm also wondering if there's an option that would allow
documents to use URIs to point to character class lists which are
appropriate to their needs.  

Parsers would have the responsibility of loading and caching these
documents as required to process documents. (The default value would
reflect XML 1.0 behavior.) XML 1.0 parsers would trip over the
declaration and raise an error, keeping these non-XML 1.0 documents out
of the way of such parsers.

Ideally, this approach would permit a RDDL-like decoupling of document
and metadata, though in this case the metadata would be rather
fundamental information.  In some ways this reinvents the SGML
Declaration, but I fear that we're well past that already.

There are some potentially bizarre results, if developers took it on
themselves to build funhouse lists where digits were letters and
combining characters digits etc., but I don't think the odds of that are
too substantial.

Simon St.Laurent
http://www.simonstl.com