I expressed my view of them in the B-spline survey %Boor76 to the effect that they also make sense for arbitrary grids, being, in effect, the B-spline coefficients of a spline with some knot sequence but expressed as a linear combination of B-splines on a finer knot sequence. %
Lyche gives error bounds for discrete cubic spline interpolation in %Lyche76c . %
Dikshit and Powar discuss `Discrete cubic spline interpolation' in %DikshitPowar82a . %
Rana describes the `Local behaviour of the first difference of a discrete cubic spline' in %Rana90 . %
Rana and Dubey consider the `Local behaviour of the deficient discrete cubic spline interpolator' in %RanaDubey9x .