In Fitting and Mendelsohn's textbook on First order modal logic, the concept of validity is defined only for sentences (i.e, formulas without free variables). How usual is this?
Doesn't it have a strange consequence, namely that $P(x) \lor \neg P(x)$ will not be treated as valid since it contains free variables, even though it is true in all models? So there are formulas which are true in all models but not valid in their sense. What is the rationale behind that?