Supernovae red shifts are fitted in a simple 5D model: the galaxies are assumed to be enclosed in a giant S^3-spherical shell which expands (ultra) relativistically in a (1+4)D Minkowski space. This model, as compared with the kinematical (1+3)D model of Prof Farley, goes in line with the Copernican principle: any galaxy observes the same isotropic distribution of distant supernovae, as well as the same Hubble plot of distance modulus \mu vs red shift z. A good fit is obtained (no free parameters); it coincides with Farley's fit at low z, while shows some more luminosity at high z, leading to 1% decrease in the true distance modulus (and 50% increase in luminosity) at z\sim 2. The model proposed can be also interpreted as a FLRW-like model with the scale factor a(t)=t/t_0; this could not be a solution of general relativity (5D GR is also unsuitable--it has no longitudinal polarization). However, there still exists the other theory (with D=5 and no singularities in solutions), the other game in the town, which seems to be able to do the job.

Galactic rotation curves and lack of direct observations of Dark Matter may indicate that General Relativity is not valid (on galactic scale) and should be replaced with another theory. There is the only variant of Absolute Parallelism which solutions are free of arising singularities, if D=5 (there is no room for changes). This variant does not have a Lagrangian, nor match GR: an equation of `plain' R^2-gravity (ie without R-term) is in sight instead. Arranging an expanding O_4-symmetrical solution as the basis of 5D cosmological model, and probing a universal_function of mass distribution (along very-very long the extra dimension) to place into bi-Laplace equation (R^2 gravity), one can derive the Law of Gravitation: 1/r^2 transforms to 1/r with distance (not with acceleration).