I argue that the widely adopted framework of stellar dynamics survived since 1940s, is not fitting the current knowledge on non-linear systems. Borrowed from plasma physics when several fundamental features of perturbed non-linear systems were unknown, that framework ignores the difference in the role of perturbations in two different classes of systems, in plasma with Debye screening and gravitating systems with no screening. Now, when the revolutionary role of chaotic effects is revealed even in planetary dynamics i.e. for nearly integrable systems, one would expect that for stellar systems, i.e. non-integrable systems, their role have to be far more crucial. Indeed, ergodic theory tools already enabled to prove that spherical stellar systems are exponentially instable due to N-body interactions, while the two-body encounters, contrary to existing belief, are not the dominating mechanism of their relaxation. Chaotic effects distinguish morphological and other properties of galaxies. Using the Ricci curvature criterion, one can also show that a central massive object (nucleus) makes the N-body gravitating system more instable (chaotic), while systems with double nuclei are even more instable than those with a single one.