Version 2.1
        o >>>>FINALLY<<<< discovered and implemented PROVED TO BE 
	  OPTIMAL dynamical aproximation algorithm in linear time!!!
	  It is still bit slow..but it is here!
	  It should slow down at the begining but certainly speeds up
	  during zoom
	  Also allows deeper zoom
	o added out of memory checks to make xaos safer on non unix platforms
        o autopilot is now timed by same way as all other parts are.
	  Now will act same on all computers(I hope)
	o Memorry corruption bug fixed - it caused problems with ysizes not dividable
	  by 2 with some other strange situations.
	o New README.interface
	   This helps others to extend XaoS or use its routines
	   in their software
	o Added many sanity checking to make easy debugging of new ports
        o small cleanups (unused variables etc..)
	o minnor speedups
	o added OS/2 port(graphics 320x200 and text) By Thomas A. K. Kjaer
	o added macintosh support by Tapio K. Vocadlo (thanks)
	o added support for different scanline size than width. for Mac and X11 port
	  >>WARNING<< changes to driver API:alloc_buffers now returns line size(usually
	  width)
	o Thomas' email updated
	o Bit less agresive to gcc optimizer..no more compiler crashes..
	  I hope
    UI Changes:
        o <- and -> stops calculation
	o driver changing disabled when just one driver available
	o status line
	o minnor cleanups
    Plan 9 changes:
        o mkfile by Scott Schwartz
	o correct timming. Thanks to Niger Roles and others on 9fans list
    MS-DOS Changes:
	o hiddes mouse at resizes - do not cause crashes anymore
    X11 Changes:
        o private colormap support
	o Fixed freed memory accessing bug..
	o added switches:
	   -private
	   -usedefault
	   -nomitshm
	o improved autodetection of best available visual-if defaultvisual!=pseudocolor
	  try truecolor first..do not use truecolor visuals on pseudocolor displays(look
	  ugly)
	  also supports 8bpptruecolor, staticcolor and grayscale(untested)
	o Fixed bug in changing number of entries that caused problems on SGI
-------------------------------------------------------------------------------
Version 2.0
        o Configure script.
	o Added biomorph outcoloring mode
	o Compiles now with -O2 - no -O3 required..
	  but kept -O6 for invoking pgcc optimizations
	o Many changes to make binary file smaller...consts, static, no
	  inlines and so on..
	  about 10KB from full XaoS saved..
	o Long doubles automatically disabled on SunOs
	o added new curses based interface. Have you ever seen realtime 
	  256 color fractal zoomer with text based output? :)
	o Now just one compact binary for all available drivers
	o Faster startup..(three times)
	o Compiles w/o many warnings(except these in libraries and
	  implicit definitions) with -ansi -pedantic -traditional -Wall
	o Autopilot not detects "oscillating" and too deep zoom so it can work
	  to infinity...can run as demo w/o any user input..
	  also auto detects major changes done by user and reinitializes variables...
	  so it restarts after changing of fractal type and so on..
	o many many new bugs!
	  But remember! Its not bug! Its feature!
   Ugly interface changes:
        o Completely redesigned-hardware depended parts are now
	  coded as drivers with common interface-easier adding
	  of new architectures
	  but looks as ugly as in previous releases :)
	o Added changing of display driver on the fly
	o Displays calculation status
	o Possibility of interrupting calculation
	o Reduced size of output program
	  (for example SVGAlib version from 48 to 37)
	o speed changing is now with displaying of status..
	o all ports modified for new interface
	o better status informations
	o new keys:
	  '?'/'/' status information
	  'r'     recalculate
	  'z'     interrupt
	  'v'     driver change
	  ' '     Redraw screen(or display current calculation status)
	o new hyperugly text browser that displays long and interesting
	  help text
	o Command line options support. Now you may configure XaoS
	  quite well..(turn on autopilot, set iterations etc..)
	o Changes to make drivers non reentrant
   SVGAlib prot changes:
	o Ui_svga is now event handled. It does not stay in wait
	  loop anymore.
	o Keys are now buffered-no more problems with disabling
	  autopilot
	o Added command line option:
	   -mode to select graphics mode
	o Now handles text in xmodes w/o seg. faults
	o Fixed possible crash after resizing
	o for compatibility with old SVGAlib keeps old non event handled
	  keyboard interface....but not recommended at all!
   PLAN9 port changes:
        o During porting to new ui interface I found nasty bug in displaying.
	  Now plan9 port is IMWHO surprisingly fast. 8c produces quite good code.
	  still crashes in higher resolution and has color problems
	o added keys [,] for changing iterations and ;,' for speed
	o Better palette handling
   MS-DOS port changes:
   	o Keys are now buffered-no more problems with disabling
	  autopilot
	o Compiled using DJGPP 2.01. So many DJGPP's bugs fixed :)
   X11 port changes:
        o Fixed some bugs that caused troubles when memory is not zeroed..
	o Added support for 16bpp, 24bpp and 32bpp truecolor visuals
	o Fixed bug that caused possible crashes before freeing images
	  (resizing etc..)
	o Removed unnecessary INLINEs in xlib.c/xlib.h ..
	  makes binary size bit smaller.
   Algorithm revisions:
        o Many areas inside set are not calculated
	o realloc_t - most often used array is now 32byte aligned..
	o Better newton calculating code-math now optimized by hand.
	o Small fixes in symmetry handling code
	  also minor speedups
	o Mandelbrot calculation rewrote into 386 assembly..
	  less than 5% speedup :((( (gcc guys:good work! :)
	  note that this is only for Intel platforms-xaos is still
	  compatible with non Intel computers :)

	  because of out coloring bugs disabled for now..
	o Completely new reallocing algorithms. Now proved to be 
	  optimal
	  but because of ugly looking results returned back to original one
	  current algoritmus is just about 20% slower than hypothetical
	  optimal one..still trying :)
	o Assembler fast inlined memcpy...also just for 386+
	o Changed solid guessing method so it is up to 20% more effective
	  in hope that no more mistakes introduced
	  startup speed on 486 was reduced from 6 seconds to 4
	  in fact it seems to make less errors than previous one :)
	o Small changes in Barnsley code
	o Minor fixes in symmetry code
-------------------------------------------------------------------------------
Version 1.2
   New features:
	o New recalculating with non rectangular pixels. (for full screen modes)
	  great for modes X :)
	o New palette with white background
	o Added coprocessor configuration into MS-DOS port - some coloring
	  modes are MUCH faster because of no more SIGFPE ignoring..
	o Added support for 1280x1024 and 1600x1200 into MS-DOS port
	o Default type changed back to double because long double is
	  non standard
	o 9x15 font changed to more compatible and better looking fixed.
	o First release of plan9 port (very interesting OS see
	    http://plan9.att.com..trial version if free)-not 
	    complete yet..colors looks ugly and keys like 
	    arrows does not work.. 

	    someone who can help me?

	o Added key defines for old SVGAlib compatibility.

   Fixes:
	o Fixed bug that caused problems on dos configuration with
	  lot of free memory :)
	o Fixed symmetry handling in binary decomposition+lambda planes
	o fixed some bugs that caused core dumps on non Intel
	    architectures
	o Floating point exception is now ignored-no more
	    problems with random overflowing...(but great slowdown
	    in cases that caused sigfpe)
	o lots of cleanups for plan9 cc.
	o Width of font is autodetected. fixed problems with
	    non-standard X servers
	o Avoided pgcc bug that caused wrong calculating of octo fractal
	o Fixed color inside set(not fixed 0)
	o Fixed color handling in X11
-------------------------------------------------------------------------------
Version 1.1
	o Removed unusable fixedpoint routines(386 users - go back to 1.0!)
	o New planes (toggled by pressing "i")
		- mu		(normal complex plane)
		- 1/mu		(inversed)
		- 1/(mu+0.25)
		- lambda
		- 1/lambda
		- 1/(lambda-1)
		- 1/(mu-1.40115)
	o Symmetry code completely redesigned
	o Added "m" into help
	o Better approximation algorithm in "mkrealloctable";
	  now it is about twice faster with a small zooming step
	o Approximation in approximation for yet better..also twice faster
	o Lots of cleanups
	o port for MS-DOG called chappi
	o Default palette is now same at all architectures
	o Some memory overflowing fixes(oops)
	o Fixed some divisions by zero
	o Number_t clean-allows deeper zoom for long doubles
-------------------------------------------------------------------------------
Version 1.0
	o First Release
-------------------------------------------------------------------------------

