Visual representations of color spaces

Without further explanations these graphics are rather dubious, in my humble opinion.

http://docs-hoffmann.de/ciexyz29082000.pdf

On p.8 we can see, how XYZ is mapped onto the plane X+Y+Z=1 as coordinate system xy
by a perspective projection:

The original Y-axis is not orthogonal to the plane xy, therefore the quoted diagrams
are not correct. It's of course possible to construct a kind of substitution Y* orthogonally
to xy.

As long as we are talking about unrelated colors, like an arbitrary set of spectral light
sources with arbitrary radiant powers, the human gamut in XYZ is indeed a cone,
which has its vertex at X=Y=Z=0 and which intersects the plane X+Y+Z=1 by this
horseshoe curve (plus purple line for mixtures of extreme spectral red and blue).

About power:

http://en.wikipedia.org/wiki/Radiant_flux
"radiometric power", correctly: radiant power or radiant flux

The situation becomes different under certain circumstances where the unlimited cone is
replaced by a finite volume or by finite curves on this cone:

1) The radiant power of the spectral light sources is limited and equal, e.g. 1 Watt.
    The curve on the cone shows the spectral loci. In Wyszecki&Stiles in chapter 3.3.3
    in Fig. 4(3.3.3) exactly this situation is illustrated: we can see the cone, the horseshoe and

    the spectral locus on the cone.
    As already mentioned, this issue has been further evaluated by Jim Worthey ("amplitude
    not left out" and illustrated by me and of course by himself  (see references):
    http://docs-hoffmann.de/jimcolor12062004.pdf

2) The volume of Optimal surface colors, concept by Roesch, as already explained.
    Fig. 5(3.7) in W&S shows Y orthogonally over xy, which is not correct, but we understand
    the meaning. This 'wet sack' has on top Y=100, which is the luminance of the illuminant.

3) A color space is a subset of a finite gamut volume in XYZ. There is always a white point and
   nothing can become brighter than the white point. An RGB space is represented in XYZ by an
   affine distorted cube. Mapping to xy delivers a triangle, but here we can see, what happens

   if the "amplitude is not left out":
   http://docs-hoffmann.de/ciegamut16012003.pdf
   The sections Y=const. with the dark sides of the cube appear still as triangles, but the sections

   with the bright sides by other polygons (p.12 for sRGB).

   A threedimensional graphic would be possible, Y over xy.

Best regards --Gernot Hoffmann