In the paper the void is modeled as a spherical underdense region surr
ounded by shells changing into the Friedmannian exterior. This model i
s in fact a single Tolman-Bondi metric, where at the edge of the void
the density need not be continuous. In principle, there may exist 72 v
ariations. These models contain also the cases, when in the void itsel
f there is a Minkowskian vacuum; the shell crossing is not excluded, t
oo. Some technical results are obtained for the Tolman-Bondi metric. U
sing them, the questions of stability and other theoretical problems a
re investigated. Some observational facts concerning the voids are als
o used. As the key result a truncation of the possible models is obtai
ned; only 14 models are physically reasonable. This means that the uni
verse is either hyperbolically expanding (this possibility strengthens
the proposition of Bonnor and Chamorro, (1990); (1991)), or there is
a shell crossing. Thus the discovery of voids is an observational supp
ort either for the open universe or for the shell crossing scenario of
galaxy formation (Meszaros, 1991), where no anisotropy of microwave b
ackground radiation is needed in a baryon dominated universe.