AB: The over 40 year old Hartman-Perdok (HP) [Hartman and Perdok (1955). Acta Cryst. 8, 49-52, 521-524, 525-529] theory for predicting crystal morphology is reconsidered. The new approach, which gives a physical foundation to the theory, is based on F faces having a roughening transition temperature higher than 0 K. The aim of this paper is to confront the field of crystal growth and in particular the classical HP theory with modern statistical thermodynamical treatments of models of surfaces of relatively simple crystal structures. It is shown that crystal faces (hkl) containing multiple connected nets with a relatively high energy content may have a very low roughening temperature and an unexpectedly high growth rate. In some cases, crystal faces become rough at 0 K because of multiple connected nets related by symmetry giving rise to symmetry roughening. The use of connected nets in the HP theory offers the possibility of extrapolating the results of statistical thermodynamical models of simple crystal surfaces to more complex crystals as encountered in practice. The role of the step free energy in understanding crystal morphology is emphasized.