[en] The author have studied by means of numerical simulations the dynamical evolution of a network of cosmic strings, both in the radiation and matter era. Our basic conclusion is that a scaling solution exists, i.e., the string energy density evolves as t^-2. This means that the process by which long strings dump their energy into closed loops (which can gravitationally radiate away) is efficient enough to prevent the string domination over other forms of energy. This conclusion does not depend on the initial string energy density, nor on the various numerical parameters. On the other hand, the generated spectrum of loop sizes does depend on the value of our numerical lower cutoff (i.e., the minimum length of loop we allow to be chopped off the network). Furthermore, the network evolution is very different from what was assumed before), namely the creation of a few horizon sized loops per horizon volume and per hubble time, which subsequently fragment into about 10 smaller daughter loops. Rather, many tiny loops are directly cut from the network of infinite strings, and it appears that the only fundamental scale (the horizon) has been lost. This is probably because a fundamental ingredient had been overlooked, namely the kinks. These kinks are created in pairs at each intercommutation, and very rapidly, the long strings appear to be very kinky. Thus the number of long strings per horizon is still of the order of a few, but their total length is fairly large. Furthermore, a large number of kinks favors the formation of small loops, and their sizes might well be governed by the kink density along the long strings. Finally, we computed the two-point correlation function of the loops and found significant differences from the work of Turok