POV-Ray

Going by your specification of the feature, that's an absolute no-go. Here's why:

• There is essentially only one colour model in which a raytracing engine can do its colour math properly: A linear spectral model. At present, POV-Ray uses a linear RGB model, which can pass as an extremely simple (3-band) spectral model. Any other 3-dimensional colour model is fundamentally unsuited for the type of operations most frequently used in a raytracer.

• For performance reasons, a raytracing engine should not be burdened with supporting different colour models interchangeably.

• Consequently, a raytracing engine should use linear RGB internally, and linear RGB only (or a spectral model with more bands in case higher colour fidelity is desired).

• Therefore, linear RGB (or alternatively a spectral model) is, and will always be, the sole colour model by which the POV-Ray render engine interfaces to the outside world. For other colour models support may be provided by means of syntactic sugar in the parser (as in the case of sRGB) or standard macros (as in the case of HLS and others), but either way they will always be converted to linear RGB before passing them on to the render engine. (Even the assumed_gamma keyword is in some sense just syntactic sugar for "screw proper gamma handling in the conversion", and is there solely for legacy reasons anyway.)

• The colour_map mechanism is indeed part of the render engine. Therefore, the colours in a colour_map will always be linear RGB behind the scenes, and no matter what colour model you originally specified the colours in, the "natural" interpolation will always be in linear RGB space. (Besides, what behaviour would you expect if you specified the colours in a single colour_map using multiple different models?)

In my book there is exatly one reasonable approach to achieve what you are trying to do: Implement a non-linear interpolation mode for colour_map (and also for pigment_map while we're at it), as a feature that is independent of the colour model originally used to specify the colours. it would explicitly take linear RGB colours, interpolate them non-linearly according to some special rule (which may or may not involve actual conversion to HSL), and explicitly produce linear RGB colours again.

As a matter of fact, the basis for such a feature already exists in POV-Ray 3.7.1 (currently in development, with source and semi-official Windows builds publicly available), in the form of the blend_mode keyword, which may already be doing what you're trying to achieve, as one of the modes available uses linear interpolation for chromaticity but non-linear interpolation for brightness.

Now tracked on github as issue #281.