We present the first dynamically and thermally self-consistent calculations of fast adiabatically expanding gas flows from the Galactic disk into the halo. It is shown that in a hot plasma (T >= 10^6 K) with a high overpressure with respect to the ambient medium, the dynamical time scale is much shorter than the intrinsic time scales (e.g. for recombination, collisional excitation and ionization etc.). Therefore dynamical models that use collisional ionization equilibrium (CIE) cooling functions for the evolution of the plasma are in general not correct. In particular, the emission spectra obtained from non-equilibrium calculations are radically different. We describe a method to obtain self-consistent solutions using an iterative procedure. It is demonstrated that soft X-ray background emission between 0.3 and 1.5 keV can be well explained by a superposition of line emission and delayed recombination of an initially hot plasma streaming away from the Galactic disk (outflow and/or winds). In addition to these local winds we also present calculations on global winds from spiral galaxies, which originate from a hot and quiescent galactic corona. We also emphasize that it is dangerous to derive plasma temperatures merely from line ratios of ionized species, such as NV/OVI, unless the dynamical and thermal history of the plasma is known.

We demonstrate that our Galactic Center, despite little evidence for the presence of a currently active nucleus, provides insight into the feeding of AGN: The observed velocity field of molecular clouds can be interpreted as tracing out the spiralling inwards of gas in a large accretion flow towards the Galactic Center (Linden et al. 1993, Biermann et al. 1993) in the radial distance range from a few parsec to a few hundred pc. The required effective viscosity corresponds well to the observed turbulent velocities and characteristic length scales. The implied mass influx provides indeed all the material needed to maintain the presently observed star formation rate at distances closer than about $100$ pc. We argue that the energy input from supernova explosions due to the high rate of star formation can feed the turbulence of the interstellar medium. This then keeps the effective viscosity high as required to feed the star formation. We suggest that this process leads to limit cycles in star formation, and as a consequence also to limit cycles in the feeding of any activity at the very center.