One of the open problems in nuclear structure is how to predict properties of finite nuclei from the knowledge of a bare nucleon--nucleon interaction of the meson-exchange type. We point out that a promising starting point consists in Dirac--Brueckner--Hartree--Fock (DBHF) calculations using realistic nucleon--nucleon interactions like the Bonn potentials, which are able to reproduce satisfactorily the properties of symmetric nuclear matter without the need for three-body forces, as is necessary in non-relativistic BHF calculations. However, the DBHF formalism is still too complicated to be used directly for finite nuclei. We argue that a possible route is to define effective Lagrangians with density-dependent nucleon--meson coupling vertices, which can be used in the relativistic Hartree (or relativistic mean field (RMF)) or preferably in the relativistic Hartree--Fock (RHF) approach. The density dependence is matched to the nuclear matter DBHF results. We review the present status of nuclear matter DBHF calculations and discuss the various schemes to construct the self-energy, which lead to differences in the predictions. We also discuss how effective Lagrangians have been constructed and are used in actual calculations. We point out that completely consistent calculations in this scheme still have to be performed.