Many fundamental questions for the understanding of polymer melts and
networks are more suitably addressed by current computer simulations t
han by experiments. The reason is that simulations have simultaneous a
ccess to the microscopic structure and the macroscopic behavior of wel
l-defined model systems. The coarse-grained models used often bear lit
tle relation to actual chemical species. This is justified by the expe
rimentally established universality of polymer dynamics and no limitat
ion for the test and development of theories which are directed at the
se universal aspects. The difficulties already encountered on this lev
el will be illustrated for entanglements between polymers which domina
te the dynamic in dense systems. For practical purposes it would, of c
oarse, be desirable to predict the characteristic length and time scal
es of experimental systems from the chemical structure of the polymer
chains. Due to the extremely long relaxation times it is impossible to
achieve this in brute-force simulations of truely microscopic models.
Systematic coarse-graining combined with a better theoretical underst
anding seem to offer a practical alternative.