We motivate the importance of understanding the kinematics and dynamics of the Milky Way stellar halo both in unravelling the formation history and evolution of our host Galaxy and in the more general context of galaxy dynamics. We present a cleaned picture of the kinematics of the smooth component of the stellar halo: we develop a method to quantify the average distance error on a sample of stars based on the idea of Schoenrich et al. (2012), but adapted so that it uses velocity information only on average. We use this scheme to construct an analytic distance calibration for Blue Horizontal Branch (BHB) field halo stars in Sloan colours and demonstrate that our calibration is a) more accurate than the ones available and b) unbiased w.r.t. metallicity and colour. We measure the rotation of the smooth component of the stellar halo with a tool-set of four estimators that use either only the l.o.s. velocities or the full 3D motion. From two samples of BHB stars from the Sloan Digital Sky Survey, we favour a non-rotating single halo. We critique conflicting results in the literature based on similar samples and trace back the disagreement (either in the sign of rotation or in the morphology of the halo) to sample contaminations and/or neglect account of the halo geometry. We propose a scheme that generalizes any isotropic spherical model to a model where the potential is axisymmetric and the distribution function is a function of the three actions. The idea is to approximate the Hamiltonian as a function of the actions with a library of quadratic fits to surfaces of constant energy in action space and then make explicit the dependence of the energy on the three actions in the ergodic distribution function. The transparency of the physics implied by the model we achieve, should make it possible to combine our spheroidal models to the f(J)-models of Binney (2010) for the disks and of Pontzen & Governato (2013) for the dark-matter halo, and obtain a complete actions-defined dynamical model of the Milky Way Galaxy.