
Version 2.11 (Jan 2006)
- Ported to Gimp-2.2 (required now)
- Modernized GUI a bit
- Removed About dialog, it does not worth the tarball size increase
- Code and Makefile clean-up


Version 2.10 (Jun 2004)
- Reordered buttons to match HIG, using Gimp stock Reset for Defaults.
- Added new random distribution type: Cubic tails (something between Cauchy
  and Uniform).
- Added About to boost my ego :o)
- Code cleanup.


Version 2.9 (Apr 2004)
- Gradient remix coloring mode.


Version 2.8 (Apr 2004)
- Fixed random seed initialization to match the new API.
- Added random_seed to PlasmaVals to remember `Randomize' gui state.
- Using GRand to get random numbers.
- Incorporated some 2.0 clean-up by William Skaggs.


Version 2.7
- A version numbered 2.7 was released by William Skaggs
  <weskaggs@primate.ucdavis.edu>, number skipped to avoid confusion.


Version 2.6 (Jan 2004)
- Changed dialog construction and main event loop to work with gimp-1.3.23.
- Some code cleanup.


Version 2.5 (Oct 2003)
- Works with gimp-1.3.21 and newer.


Version 2.4 (Jan 2003)
Ported to use the new Gimp random seed widget (great!).
Bug fixes:
- Tilable -> tileable.


Version 2.3 (Oct 2002)
Ported to Gimp-1.3/Gtk-2.0:
- Renamed and retyped stuff as necessary.
- Created a brand new preview widget using GtkImage & GdkPixbuf.
- Stockized stock items and reordered buttons to Gtk-2.0 order.
  Can't use GIMP_STOCK_RESET, becuase its label associates `Defaults' while
  its image `Revert'.
- Added accels to labels.  FIXME: Ugly Gimp random seed doesn't work.
- Removed Esc-exits accel, since Gtk-2.0 does it itselfs.


Version 2.2 (Sep 2001)
Internal changes:
- Simplified comitting, now uses gimp_pixel_rgns_...() functions instead of
  handling the regions by hand (still needs some cleanup).
- Uses it's own accelerator group instead of default.
Bug fixes:
- Fixed checker array overflow in plasma2_add_transparency().


Version 2.1 (Sep 2001)
Improvements:
- Generator and scaling methods
- Buttons to reset/revert settings
- Possibility to directly map to a gradient (much more smoothly since it's
  done when values are still floats)
- Grayscale can be generated even in RGB drawable
- New dialog layout
- It's faster
Bug fixes:
- More standard widget spacing
- Escape exits the dialog
Bugs:
- Generator uses gfloat for intermediate results (eating even more
  memory...)
- It shows -1 as random seed when seed is time (FIXME: how to fix it
  keeping the otherwise nice and consistent seed behaviour?)
- Cannot handle gradients whith names longer than GRADIENT_NAME_SIZE
  (this is a GIMP problem)
- Cannot reset/revert gradient (gimp exports no function for this and doing
  it by hand is obscure and error prone)
  (this is a GIMP problem)


Version 2.0 (Mar 2001)
Branched from plasma.c (not much code reused).
Note: The number of parameters changed thus it is not possible to simply
replace the old Plasma with this one.
Improvements:
- Rewrote the engine: staightforward algorithm which appeared in Fractint 19
  is used (more precisely, my implemetation of this approach from 1997).
- Added horizontal/vertical tileability options.
- Added selfsimilar (exponential, fractal, etc.) algorithm option (so it
  can really generate the thing known as `the plasma fractal').
- Some effort put into invariant behaviour (compare e.g. tileable and
  non-tileable (or selfsimilar and non-selfsimilar) images generated from the
  same seed).
- Speed optimizations.
Bugfixes:
- Fixed preview not showing final image when seed is time.
- Fixed using two-parametric plasma() as one-parametric signal callback.
- Only really needed channels are generated (3x faster on grayscale).
- Progress bar really shows progress (the old one was cheating).
- Preview correctly shows grayscale image in grayscale mode.
Bugs:
- This plugin now uses  xsize x ysize  bytes of memory and no tile cache.
  Ban me, kill me, let me administrate MS-Windows(TM), but I will still
  maintain the point it does need random access to the whole selection, so
  I don't pretend it doesn't, and even if in some very unfortunate cases
  the extra memory allocation can really be a problem, it normally makes
  the plug-in run considerably faster (note in usual RGBA case, we allocate
  only 1/4 of layer size).

