This file describes the changes made to versions of 'same'.

Changes made by Adewale Oshineye for 2.1
========
 * Replaced the reader pooling/virtual file system idea with simpler code that just loads
	all the files into memory and processes them there. Trading increased memory usage
	for reduced io improved scalability especially when working with large amounts of files.
 * Files that are smaller than the fragment size are filtered out before processing begins
 * Files in sub-folders called ".svn" or "CVS" are filtered out as they're usually not relevant.
 * Significant speed improvements when working with large numbers of files or small fragment
	.sizes


Changes made by Adewale Oshineye for 2.0
========
 * Same now takes a folder as well as a list of files

 * Various refactorings, reformatting and renaming of classes and variables

 * Set a limit on the SourceFileSystem class so that it can handle very large 
	source trees. Without the limit Same starts throwing exceptions because there
	are too many open files. Now the only limits on performance are cache size 
	(which is related to the size of the heap) and the maximum number of open files
	allowed by the operating system.

 * Removed some of the datastructures used for generating fancy stats in a bid to 
	reduce memory usage
	
 * Eliminated redundant loops
 
 * [Fixed bug]: some duplicate parts are shown more than once.
========

Version 1.2: First open source version.

 * Renamed tool from 'duploc' to 'same', because a tool named 'duploc' with
   similar functionality already exists.
 
 * Reorganized the package structure.  The entry point for the command line
   interface is now same.textui.SameCLI.

========

Version 1.1: Some bug fixes, refactoring.

 * Added the 'none' filter, which is now also the default.  This does not
   filter the input at all.

 * Added some more statistics about the matches.

 * Fixed a bug that caused matches to be skipped sometimes.  Unfortunately
   performance suffered.  But:

 * Fixed a bug that caused many comparisons to be done more than once.
   That made performance come back to an acceptable level.

 * Fixed a bug that caused the 'number of filtered lines' to be too
   high.

 * Now uses a 64-bit hash code instead of a 32-bit one.  Does not seem
   to reduce the number of potential matches significantly.

 * Moved all classes from the default package to com.baan.tools.duploc
   and com.baan.tools.test.duploc.

 * Separated the end-user I/O from the functionality.  The entry point
   is now com.baan.tools.duploc.DupLocCLI (was previously DupLoc).
   This step makes it possible to create a new front end, which
   receives events through DupLoc.IListener.

========

Version 1.0: Initial published version
