Rethinking QNames

From the "if you can't beat 'em, join 'em" department:

How catastrophic would it be if we reconsidered the definition of
QNames along the following lines:

1. Specify that the local part of a QName is NMTOKENS.

2. Specify that a QName x:y can be transformed into a URI by appending
   "y" to the URI associated with the prefix "x".

3. Specify that only a subset of QNames can be used as element and attribute
   names in XML because the local-part of an element or attirbute name
   must be an NCName.

One could go so far as to say that the prefix was also NMTOKENs, but I think
that would be going too far. I don't think we want 1:1 or 4:6 to become
QNames.

I think we might also want to consider the slighly more complicated
rule that the local part of a QName can be NMTOKENS iff it has a
prefix; unprefixed QNames must still be NCNames. That way isbn:1234 is
a QName, but 4567 is not.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | The art of being wise is the art of
http://nwalsh.com/            | knowing what to overlook.--William James