We introduce the notion of definite inequality constraints involving monoto
ne functions in a finite meet-semilattice, generalizing the logical notion
of Horn-clauses, and we give a linear time algorithm for deciding satisfiab
ility. We characterize the expressiveness of the framework of definite cons
traints and show that the algorithm uniformly solves exactly the set of all
meet-closed relational constraint problems, running with small linear time
constant factors for any fixed problem. We give an alternative technique f
or reducing inequalities to satisfiability of Hem-clauses (HORNSAT) and stu
dy its efficiency. Finally, we show that the algorithm is complete for a ma
ximal class of tractable constraints, by proving that any strict extension
will lead to NP-hard problems in any meet-semilattice. (C) 1999 Elsevier Sc
ience B.V. All rights reserved.