In this paper, we discuss the nature of the static and kinetic friction, and of (thermally activated) creep.We focus on boundary lubrication at high confining pressure (∼1GPa), as is typical for hard solids, where one or at most two layers of confined molecules separates the sliding surfaces. We find in most of our Molecular Dynamics (MD) simulations (at low sliding velocity), that the lubricant molecules are permanently attached or pinned to one of the solid walls.We describe the (flexible) lubricant-wall bonds as springs with bending elasticity. If the springs are elastically stiff, the system exhibits a very small static friction, and a (low velocity) kinetic friction which increases with increasing sliding velocity. On the other hand, if the springs are soft enough, strong elastic instabilities occur during sliding, resulting in a large static friction force Fs, and a kinetic friction force Fk equal to the static friction force at low sliding velocities. In this case rapid slip events occur at the interface, characterized by velocities much higher and independent of the drive velocity v. In the MD simulations we observe that, for incommensurate systems (at low temperature), only when the lubrication film undergoes a phase transformation at the onset of slip do we observe a static friction coefficient which is appreciately larger than the kinetic friction coefficient. We give arguments for why, at very low sliding velocity (where thermally activated creep occurs), the kinetic friction force may depend linearly on ln (v/v0), as usually observed experimentally, rather than non-linearly [−ln (v/v0)]2/3 as predicted by a simple theory of activated processes. We also discuss the role of elasticity at stop and start. We show that for "simple" rubber (at low start velocity), the static friction coefficient (?s) is equal to the kinetic friction coefficient (?k). In general, at non-zero temperature, the static friction coefficient is higher than the kinetic friction coefficient because of various thermally activated relaxation processes, e.g. chain interdiffusion or (thermally activated) formation of capillary bridges. However, there is no single value of the static friction coefficient, since it depends upon the initial dwell time and on rate of starting.We argue that the correct basis for the Coulomb friction law, which states that the friction force is proportional to the normal load, is not the approximate independence of the friction coefficient on the normal pressure (which often does not hold accurately anyhow), but rather it follows from the fact that for rough surfaces the area of real contact is proportional to the load, and the pressure distribution in the contact areas is independent of the load.