Can one improve the Froissart bound?

We explain why we hope that the Froissart bound can be improved, either qualitatively or, more likely, quantitatively, by making a better use of unitarity, in particular elastic unitarity. In other instances (Gribov's theorem) elastic unitarity played a crucial role. A preliminary requirement for this is to work with an appropriate average of the cross-section, to make the problem well defined. This is possible, without destroying the Lukaszuk--Martin bound.