We argue that the most commonly used models for nuclear scattering at ultra-relativistic energies do not treat energy conservation in a consistent fashion. Demanding theoretical consistency as a minimal requirement for a realistic model, we provide a solution for the above-mentioned problem, the so-called 'parton-based Gribov-Regge theory'. In order to keep a clean picture, we do not consider secondary interactions. We provide a very transparent extrapolation of the physics of more elementary interactions towards nucleus-nucleus scattering, without considering any nuclear effects due to final state interactions. In this sense we consider our model a realistic and consistent approach to describe the initial stage of nuclear collisions.