2006-06-27:: Version 2.6
* Removed NilClass#method_missing.  I have replaced it with a new class
  called KBNilClass.  Thanks to a host of people for help on this,
  including:  Assaph Mehr, James Edward Gray II, Matthew Desmarais,
  Logan Capaldo, Trans, John Carter, Dave Burt and anyone else I missed.
* Added conditional require logic so that KirbyBase will use FasterCVS if
  it is available.  Thanks to James Edward Gray II for this.
* You can now delay index creation in local mode.  Thanks to Nicholas Rahn
  for this.
* Added ability to allow for a custom record class with no kb_create or
  kb_defaults methods.  KirbyBase will return each result record as an
  instance of the custom record class, and will attempt to set attributes
  with the same names as the table's field names equal to the values of
  the table record's fields.  Thanks to Hal Fulton for this idea.

2005-12-30:: Version 2.5.2
* Changed the behavior of KBTable#insert method.  If user specifies nil
  for a field value and there is a default value for that field, the
  default value will no longer override the user specified nil value.
  Thanks to Assaph Mehr for suggesting this.

2005-12-28:: Version 2.5.1
* Fixed a bug that had broken encrypted tables.
* Changed KBTable#pack method so that it raises an error if trying to
  execute when :connect_type==:client.
* Fixed a bug where it was possible to insert records missing a required
  field if using a hash.  Thanks to Adam Shelly for this.
* Fixed a bug that occurred when you tried to update records using a
  block and you tried to reference a field in the current record inside
  the block.  Much thanks to Assaph Mehr for reporting this.
* Fixed a bug that allowed you to have duplicate column names.  Thanks to
  Assaph Mehr for spotting this.
* Changed the way KBTable#set works with memo/blob fields.
* Started creating unit tests.
* Changed the KBTable#clear method to return number of records deleted.
  Thanks to Assaph Mehr for this enhancement.
* Moved #build_header_string from KBEngine class to KirbyBase class.
* Added KirbyBase::VERSION constant.

2005-12-01:: Version 2.5
* Fixed a subtle bug in KBTable#create_indexes.
* Added the following new methods to KBTable:  add_index, drop_index,
  rename_column, change_column_type, change_column_default_value, and
  change_column_required.
* Added the ability to specify a default column value at table creation
  time.
* Added the ability to specify, at table creation time, that a column value 
  is required when inserting or updating records.
* Removed #add_table_column and #drop_table_column from KirbyBase class
  and added #add_column and #drop_column to KBTable class.  I felt like
  it made more sense to have these methods in the table's class rather
  than the database's class.
* Added KirbyBase#rename_table method.
* Added the ability to, upon database initialization, specify that index
  creation should not happen until a table is actually opened.  This
  speeds up database initialization at the cost of slower table
  initialization later.

2005-11-13:: Version 2.4
* Added a new column type: :Time.  Thanks to George Moschovitis for coding
  this enhancement.
* Added more functionality to Memo and Blob fields.  They are no longer
  just read-only.  You can now also write to them from KirbyBase.  The
  interface for Memo and Blob fields has changed because of this.
* Added the ability to specify, when you initialize a database connection,
  a base directory where memo/blob fields will be stored.
* Changed the way indexes are handled by KBTable in client/server mode.
  Now, when KBTable grabs an index from KBEngine, it will hold onto it and
  re-use it unless it has been modified since the last time it grabbed it.
  This speeds up subsequent queries on the same index.
* Removed the restriction that the child table had to exist before you
  could define a Link_many field in #create_table.  I did this so that
  it would possible to now define many-to-many links.  See the example in
  the distribution.  This also goes for Lookup fields.
* Added two sample scripts:  kbserverctl.rb and kbserver_daemon.rb, that
  show how to set up a KirbyBase server process as a Windows Service.
  Thanks to Daniel Berger for his excellent package, win32-service.
* Thouroughly revised the manual.  I used the excellent text document
  formatter, AsciiDoc.  Many thanks to Stuart Rackham for developing this
  great tool.
* Fixed a bug in KBTable#clear that was causing the recno counter not to
  be reset.  Thanks to basi for this.

2005-10-10:: Version 2.3 
* Added ability to specify lookup fields.
* Added ability to specify one-to-many links between tables.
* Added ability to specify calculated fields in tables.
* Added Memo and Blob field types.
* Added YAML field type.  Many thanks to Logan Capaldo for this idea!
* Added indexing to speed up queries.
* Two new methods:  #add_table_column and #drop_table_column.
* Added the ability to designate a table field as the "key" field, for
  Lookup purposes.  This simply makes it easier to define Lookup fields.
* Added "crosstab" capabilities to KBResultSet.

2005-08-09:: Version 2.2.1
* Fixed a bug in with_write_lock.  Thanks to Zed A. Shaw for this bugfix.
* Fixed a bug that occurred if @record_class was a nested class.  Thanks
  to Hal Fulton for this bugfix.

2005-05-02:: Version 2.2
* By far the biggest change in this version is that I have completely
  redesigned the internal structure of the database code.  Because the
  KirbyBase and KBTable classes were too tightly coupled, I have created
  a KBEngine class and moved all low-level I/O logic and locking logic
  to this class.  This allowed me to restructure the KirbyBase class to
  remove all of the methods that should have been private, but couldn't be
  because of the coupling to KBTable.  In addition, it has allowed me to
  take all of the low-level code that should not have been in the KBTable
  class and put it where it belongs, as part of the underlying engine.  I
  feel that the design of KirbyBase is much cleaner now.  No changes were
  made to the class interfaces, so you should not have to change any of 
  your code.
* Changed str_to_date and str_to_datetime to use Date#parse method.
  Thanks to Emiel van de Laar for this enhancement.
* Changed #pack method so that it no longer reads the whole file into
  memory while packing it.
  Thanks to Hugh Sasse for reporting this bug.
* Changed code so that special character sequences like &linefeed; can be
  part of input data and KirbyBase will not interpret it as special 
  characters.
  Thanks to Hugh Sasse for this bug fix.
    
2005-04-11:: Version 2.1
* Changed the interface to KirbyBase#new and KirbyBase#create_table.  You
  now specify arguments via a code block or as part of the argument list.
* Added the ability to specify a class at table creation time.  
  Thereafter, whenever you do a #select, the result set will be an array
  of instances of that class, instead of instances of Struct.  You can
  also use instances of this class as the argument to #insert, #update,
  #set.
* Added the ability to encrypt a table so that it is no longer stored as
  a plain-text file.
* Added the ability to explicity specify that you want a result set to be
  sorted in ascending order.
* Added the ability to import a csv file into an existing table.
* Added the ability to select a record as if the table were a Hash with 
  it's key being the recno field.
* Added the ability to update a record as if the table were a Hash with
  it's key being the recno field.

2005-03-28:: Version 2.0
* New version with completely new interface.  A huge thanks to Hal Fulton
  for all of his help with this new version.
