Put the contents of the README into the documentation. 
  First, we need to distribute a pre-compiled documentation

xml module:
============

* DOM requires Utf16-encoding for strings.
  Can we make all the packages generic, so that the user can decide whether
  to use Utf16 or Utf8 ?

* Unicode: Add conversion to Wide_Char and Wide_String

* Support for xml:* attributes (space, ...)
  Note that xml:space means that we must use the Ignorable_Whitespace callback
  in SAX.

* Should Encoding_Scheme include support for From_Utf32 and To_Utf32

* Unicode: http://www.unicode.org/unicode/standard/versions/Unicode3.0.html

* DOM: Finish implementation for the few remaining cases

* SAX: Should have a feature to specify whether entity refs in attribute
  values should be expanded or not

* Parser: Checking of self-contained entities could be cleaner (ie done at
  a more general level). For instance, directly at the end of Next_Token,
  but we must exclude Text from the tests.

* Can we implement the lexical parser as a state machine (Next_Char called
  at a single place). This might make it shorter (but probably not more
  efficient)

Release 0.7
===========
* Parse_Element_Model_From_Entity should use Next_Token (for entitites)
* File_Input shouldn't store the whole file in memory (have a buffer of 64k
  instead).
* No need to return Id in Next_Token, should be stored directly in
  Parser.Id. This would be more efficient. This allows for a simplification
  in Parse_Element_Model.
* For efficiency, Sax.Attributes should be an array
* We must normalize PublicID (see 4.2.2)
* Group all DTD-related field into a separate record (Entities, Default_Atts)