This version
============

* Create new dispatcher example script.

* Add it to docs and index.html.

Future versions
===============

* When reading entity and param tags, what is hash val when we've seen an
  external id (and possibly an NDATA)?

* Ensure that xml preprocessing instruction only sees "version",
  "encoding", and "standalone" tags and no others.

* Ensure that xml preprocessing instruction attributes are seen in the
  proper order. Warning; not error.

* Add test code for

	* Eight-bit tag names (new NAME_REGEX).
	* "encoding" and "standalone" attrs of xml preproc instruction,
	  including bad values.
	* Illegal DOCTYPE SYSTEM and PUBLIC formats.

* Better writer prettifying code.

* Should I stop saving XML source into entities completely? What good does
  it do, except save a few milliseconds for Writer objects that are writing
  Document objects?

* Add all XML spec well-formedness checks. See list in README file
  LIMITATIONS section.

* Everything else listed in the README file LIMITATIONS section.

* Make more efficient. (Process already started.) I'm probably creating too
  many input objects in the tokenizer.

* Supply default attribute values. (Have to start paying attention to
  ATTLIST tags first, of course.)

* Comment code, including classes and methods.

* Redo tokenizer:

    * Change tokenizer to slurp lines or characters as needed (if input is
      not a string). If I do, the argument to the tokenizer constructor
      will need to respond to more than just :read (e.g., :eof?, :seek).

    * How much do I care that the entire XML string is in memory? Strings
      passed to me are probably small enough, but file names may point to
      huge files. I could rewrite the tokenizer to use IO objects instead
      (for example, either use the file passed to me or write the XML
      string to a temp file). If I decide not to make this change, the next
      item is not really necessary.

    * Abstract Tokenizer's peek*, nextChar*, line, column, and pos methods,
      etc. into a ReadStream class. If I do this, the tokenizer will have
      to change a bit because it assumes it has access to the XML as a
      single string.

Rejected ideas
==============

* Add methods to Node that allow searches for children with specified name?
  Nah, this isn't necessary; just iterate over node.children and check for
  the name you want. For example (from the test code):

    node = doc.rootNode.children.detect { | n |
	n.entity.instance_of?(NQXML::Tag) && n.entity.name == 'simpleTag'
    }

Notes
=====

No start-tag, end-tag, empty-element tag, element, comment, processing
instruction, character reference, or entity reference can begin in one
entity and end in another.