How can star formation be sustained?

There is overwhelming evidence that the Milky Way has formed its stars at a relatively constant rate throughout the Hubble time. This implies that its stock of cold gas was not in place since the beginning but it has been acquired slowly through gas accretion. The gas accretion must have been at low metallicity in order to reconcile the metallicities observed in the disc with the predictions of chemical evolution models. But how does this gas accretion take place? I review the current evidence of gas accretion into the Milky Way and similar galaxies through the infall of cold gas clouds and satellites. The conclusion from these studies is that the infalling gas at high column densities observed in HI emission is a least one order of magnitude below the value required to sustain star formation, thus accretion must come from a different channel. The likely reservoir for gas accretion is the cosmological corona of virial-temperature gas in which every galaxy must be embedded. At the interface between the disc and the corona the cold high-metallicity disc gas and the hot coronal medium must mix efficiently and this mixing causes the cooling and accretion of the lower corona. I show how this mechanism reproduces the kinematics of the neutral extraplanar gas in the Milky Way and other nearby galaxies and the ionised high-velocity clouds observed in HST spectra. I conclude with the speculation that the loss in efficiency of the disc-corona interaction is the ultimate cause for the evolution of disc galaxies towards the red sequence.