[en] The Dirac impulse approximation has to date provided dramatic improvement in our ability to predict, with no free parameters, spin observables in proton-nucleus elastic scattering at intermediate energies. The key ingredients of this approach are Dirac propagation and the nucleon-nucleon invariant amplitudes. So far, local approximations to the NN amplitudes have been used. The standard NN representation in terms of Dirac scalar, vector, and so on, parts which is free of kinematical singularities seems to naturally predict the correct coupling to negative energy states for energies above 300 MeV. At low energy, this coupling is subject to an ambiguity between pseudoscalar and pseudovector πN coupling mechanisms and it is evident that the pseudoscalar coupling treated in a local approximation causes too much scalar-vector difference and thus too large pair contributions. Once this problem is remedied, the Dirac optical potential is expected to be calculable from a nucleon-nucleon quasi-potential over the range 0 to 1000 MeV. For the energy region above about 300 MeV, the large scalar and vector potentials of Dirac phenomenology are seen to be accurately predicted by the impulse approximation. Work by Shakin and collaborators provides complementary results at low energy based on a nuclear matter g-matrix. A basic conclusion is that relativistic spin effects cannot be neglected in nuclear physics. 36 references