[relax-ng] Some ideas about type-assignment

[EB2M-MRT@asahi-net.or.jp (Murata Makoto)]

- Type-assignment should be useful for XQuery.

- Type-assignment should be useful for data binding tools.

- Type-assignement should be separated from validation (It should be
  possible to do type-assignment without performing validation, and 
  vice versa)

- Ideally, deep understanding of the RELAX NG spec (and underlying
  hedge automata) should not be required for implementing
  type-assignemnt.  (It might be an idea to create XPath 
  expressions from RELAX NG schemas.)

- Type-assignment does not have to be possible for every RELAX NG 
  schema.  (When it is not possible, diagnostic message should 
  be issued.)

- Given a RELAX NG schema and a document, type-assignment should
  
  1) uniquely determine the datatype (data/text/value/list) of
     each string (i.e. attribute values and mixed contents),

  2) uniquely determine the named pattern for every element,

  3) uniquely determine which group/interleave/choice is applied
     to the content of each element and the value of each attribute, 
     and

  4) do (1) for each token created by <list>, and do (3) for 
     each token sequence

  5) determine if whitespace is ignorable

  Note: I am not sure which of them should be adopted.

- Type-assignment should not impose much burden on schema authors.