In this paper, we propose a variant of stable model semantics for
disjunctive logic programming and deductive databases. The
semantics, called minimal founded, generalizes
stable model semantics for normal (i.e. non-disjunctive) programs,
but differs from disjunctive stable model semantics (the extension
of stable model semantics for disjunctive programs). Compared with
disjunctive stable model semantics, minimal founded semantics seems
to be more intuitive, it gives meaning to programs which are
meaningless under stable model semantics and is no harder to
compute. More specifically, minimal founded semantics differs from
stable model semantics only for disjunctive programs having
constraint rules or rules working as constraints. We study the
expressive power of the semantics, and show that for general
disjunctive datalog programs it has the same power as disjunctive
stable model semantics.