Four features of the interaction between the satellite Io and the surrounding jovian magnetosphere place constraints on the atmosphere of the satellite and on the nature of the interaction: (1) the atmosphere must be at least partly an exosphere so that sodium and other non-volatile atoms sputtered from the surface of the satellite could escape1,2; (2) the atmosphere should be dense enough for a detectable ionosphere to form showing a leading-trailing asymmetry2,3; (3) the atmosphere should provide an electrically conductive path to produce the electric currents which eventually cause the Io-related decametric radio emissions4,5; (4) the conductivity of the ionosphere should not be large enough to short out the electric field seen by the satellite, as pronounced absorption of energetic particles is observed6-8. Before the Pioneer 10 encounter it was pointed out9 that erosion by the co-rotating magnetosphere caused atmospheres on the galilean satellites with surface pressures in the range 10-5-10-9 mbar, if they exist, to be stable only if they are continuously replenished and therefore likely to be associated with "venting or the presence of non-hydrous ices on the surface". Alternatively, if the surface is covered with non-volatiles and/or water ice, the atmospheric pressure would be very low and controlled by sputtering due to the impact of magnetospheric plasma and energetic particles on the surface 9. In any case, the interaction with the magnetosphere is likely to be 'weak' as it is for Mercury10, that is, the magnetic field and plasma sweep into and past the satellite being affected mainly by the additional inertia of newly formed ions (provided the satellite is non-conducting and non-magnetic). We substantiate these arguments here using the recent Voyager observations, and suggest that the interaction between Io and the magnetosphere can be understood if the atmosphere of Io were largely exospheric in nature and controlled by the vapour pressure of SO2 ice which covers the surface.