TODO (1.9)
==========

* Made <assign> statement an optional block element, a-la JSTL.
  - Update the manual to reflect this behaviour.

* Function variables can be either globally or dynamically scoped.
  - Update the manual to reflect this behaviour.

* Similarly, functions take on Awk behaviour regarding local variables.
  - Update the manual to reflect this behaviour.

* Added an #empty special identifier, which returns a null model.
  - Update the manual to reflect this behaviour.

* Improved the way the reflection library deals with numbers and number
  conversions from the internal template models.
  - Add more test cases for this model.

* Check the severity of the new exception handler calls. They may be a bit out.

* Consider a FastSubList model, that takes an argument of the form [ 1 .. 10 ]
  and returns a TemplateListModel2 corresponding to the numbered arguments. The
  arguments would be assigned meaning by the underlying model. Todo: decide
  if the sublist should also be indexed, and if so, what the indexed should be.

* Created a rewrite of the freemarker.ext.beans classes (beans2).
  - Need to extensively document the behaviour of the new package.

* FastIndexedIterator and FastIndexedWriteableIterator could implement
  TemplateHashModel and provide an "index" variable for the current position
  in the list.

* Also consider making FastListIterator implement indexed model. Add a
  constructor to take the initial index offset. Similar for EnumIterator.

* Lots of JavaDoc needs to be written. Much has been neglected during this
  development cycle.

* Similarly, the file template needs its copyright statement updated to 2005.

* Look at AtomicBoolean, AtomicInteger, AtomicLong, and the
  java.util.concurrent.atomic package generally, and see what can be handled
  within the reflection package.

* Look closely at the javax.script package in Java 1.6 draft, and see if there
  is a suitable adapter we can create, or remodel existing classes on.

TODO 1.10
=========

* Consider regularising indexes to be 1-based, rather than 0-based. This makes
  more sense given the engine's list range notation.

* Along similar lines, can finally make the switch to relative paths for
  handling inclusions. This may further complicate IncludeInstruction and/or
  the cache package.
  - How do you resolve non-filesystem paths? A pluggable resolver? Where do you
    plug it in?

* Have yet another look at making StandardTemplateParser use a Reader instead
  of String and parsePos, etc. Need a markable buffer plus a small lookahead
  buffer.

* Models deserve their own package. Consider creating one especially for the
  model interfaces.