Baryons seem to dominate the central parts of galaxies (within 1-2 Re), probably even in dwarfs, and stellar mass-to-light ratios are consistent with the dynamical estimates in these regions. The mass discrepancy increases outwards, which suggests the presence of dark matter, including in early-type galaxies. The outer mass profiles are roughly consistent with NFW profiles, but these seem to fail to reproduce the central, shallow, density slopes.
Each technique mentioned above represents one path to the determination of mass profiles in individual galaxies. They all have their own specific (nagging) issues, and include many geometrical, physical, numerical assumptions which should be carefully examined in turn. As emphasised, point-like sources, background flux, counts and the assumption of hydrostatic equilibrium have to be considered when using X-ray spectroscopy and imaging. Strongly lensed systems are rare, and biased toward "relatively" high redshifts and large masses. HI, H, and CO are clearly complementary, but we always need to examine large-scale perturbations, as well as the assumptions we make on the gas dynamics. Outer gas rings are efficient tools in this context but only probe the potential locally, and are, unfortunately, very rare. Density waves such as bars, spirals and warps are good tracers of the dynamics of the system (hence of its mass), but may not bring very strong constraints, and are usually restricted to the central regions. Stellar dynamics is complex and require state-of-the-art modelling with well tuned assumptions. Still, degeneracies exist which may sometime blur the robustness of the results. Tracers such as planetary nebulae, globular clusters or streams are obvious targets, but it is not clear if these are effectively good tracers and if basic assumptions such as stationarity can be applied. Finally M / L derived from stellar populations is probably the least robust parameter provided by spectral synthesis models, although we witness steady progress there. More generally, I think we can be relatively optimistic as we today follow the advent of new instruments, techniques, and numerical codes which clearly address the issues mentioned here. We then need to find out if such vastly different approaches deliver consistent answers when probing the same mass regimes.
I would like to thank the organisers for providing me with the opportunity to participate to such a fruitful conference. I would also like to warmly thanks colleagues who helped me prepare this review, namely Michele Cappellari, Roelof de Jong, Harald Kuntschner, Stacy McGaugh, Gary Mamon, Claudia Maraston, Tom Statler and Anne-Marie Weijmans.