version 4.4 (2010-01-04) Made it possible to restrict sql authentication to
	one ip only. Moved the image conversion to separate package and
	merge the different image handlers into just one ImageHandler that
	uses the right image converter internally. Made ImageHandler fall
	back to java based image conversion if the external process
	fails. Merged the different proxy authentication filters into one
	filter class and several backends. Moved proxy authentication
	caching up to ProxyAuth so that all authenticators can use it.
	Added some logging to the image handler. Setting repack to true and
	compress to false now works. Added a setting "allow_without_auth"
	to the ProxyAuth so that it possible to access some pages even if
	proxy authentication is generally required. Fixed a problem with
	HTTP/0.9 request.

version 4.3 (2009-11-17) Some code cleanups identified by findbugs.
	Changed the handler setup to use full regexps. Updated the handler
	setup so that it catches more images. Updated the sample web
	server so that it is more secure and sends correct mime type in
	more cases. Updated dnsjava to 2.0.7. Added a image handler that
	uses ImageIO/ImageWriter so it is full java and that means that it
	does not have to fork a convert process. Updated some filter
	documentation. Made it possible to configure SQLProxyAuth to cache
	the results.

version 4.2 (2009-08-20) Made it possible to specify what ip rabbit
	should listen on. Made it possible to specify what ip rabbit
	should use for outgoing connections. Made access logging handle
	log setup failure without getting NullPointerException.  Added
	task timing and a status page that can show the time take by
	different tasks. Do not log about empty cache dir on startup.

version 4.1 (2009-07-01) Increased the default keepalive timeout to 30
	seconds. Made number of selector threads configurable, with a
	default value of the number of cpu cores. Made selector status
	page easer to read by altering colors between selectors. Made
	WebConnectionResourceSource actually grow the download buffer when
	it has read full chunks a few times. Made ConnectionHandler use
	concurrent collections for better scalability. ConnectionHandler
	now removes all empty pools. Tried to fix a NPE in the selector
	code. Call Deflater.end to release native zip resources earlier.

version 4.0 (2009-05-03) Reworked nio layer so that rabbit uses multiple
	selector threads. Reworked logging to use java.util.logging and
	made it possible to use either rabbit.conf or the system property
	java.util.logging.config.file to specify the logging
	configuration. Fixed a problem where web connections could be put
	into the connection pools multiple times.
