With absolutely
NO knowledge whatsoever, I will take a poke in the dark.
Since in Java,
URI is not a subclass of String, it make sense that anyURI
in XSD should not be a subclass of String.
My argument would be: go to the definition of URI. RFC 3986 says
"A Uniform Resource Identifier (URI) is a compact sequence of
characters that identifies an abstract or physical resource."
That says a URI is a string with certain properties. So it should be
modelled as a string in the schema (and should therefore inherit the
properties and operations applicable to other strings). Just as
xs:NCName is a string, so should xs:anyURI be.
That makes URIs different from dates, or integers, for example. One
can argue that dates are not strings. If you go looking for a
definition of dates, or integers, I think it's unlikely to say that
dates/integers are sequences of characters.
But I have a lot of sympathy with Jeni Tennison's proposal for data
types which says that everything is a string...
