What are the proposed rules for exception signature compatibility
for redefinitions of virtual functions?  I would *hope* that a redefinition
in a derived class would be required to have the same exception signature
as the declaration in the base class, just as the function's return type
must be the same.  Theoretically I could even see the redefinition being
allowed to have a "subset" signature, just as (proposedly) the return
type can be a sub-type.