I am planning to simplify gsl histograms to use linear scales in their coordinates, dropping support for nonlinear scales (the observations can be transformed nonlinearly instead). If anyone sees a need to keep nonlinear histogram scales let me know.