only take up minuscule amounts of time. Therefore we're not goint to
      worry about them.

     early bailing out)

proactively. Fixed the condition to remove scalars only when clip-by-scalar is
on and "GenerateClipScalars" is off.

ENH:Update glext.h from version 51 (2009/04/29) to version 52 (2009/05/05) from http://opengl.org/registry

ENH:Draft version of OpenGL 3.1 header file (2009-05-05 15:53:39 -0700) from http://www.opengl.org/registry

composite datasets with partial arrays.

     debug mode

ENH: Updated to build boost from the Boost SVN trunk rather than from the PBGL SVN repository since PBGL is now in Boost proper.

ENH: Add support for 64-bit integers and transferring vtkIdType (including between processes with different widths).

     vtkTimerLog.

version of the filter. Fixed that. Also fixed vtkPExtractArraysOverTime to work
correctly when the "multiblock-of-tables" produced by vtkExtractArraysOverTime.

 - Also removing "vector-of-vectors" support from vtkAnnotationLayers.
   vtkAnnotationLayers is now a simple collection of annotations, plus
   an optional current selection object. This is the first step toward
   combining vtkAnnotationLink and vtkSelectionLink.

ENH: Made a correction to the RGB to XYZ color transfer (I made some missinterpretations about gamma correction and linearity in physical and perceptive).  This had several ramifications for CIELAB color spaces and transfer functions that rely on it.

a) Tracks all scalars present on the input data.
b) uses an Id channel array if specified.
c) Filter now outputs 2 datasets, the first is the trails, the second if the lead point on the trail. This is useful if you glyph the lead point and tube the trail, nice images can be generated.
d) much more stable in general
Many thanks to Ken Moreland for contributions to this class and to Benjamin Schindler of ETHZ for extensive testing and suggestions.