Local consistency has proven to be an important concept in the theory
and practice of constraint networks. In this paper, we present a new d
efinition of local consistency, called relational consistency. The new
definition is relation-based, in contrast with the previous definitio
n of local consistency, which we characterize as variable-based. We sh
ow the conceptual power of the new definition by showing how it unifie
s known elimination operators such as resolution in theorem proving, j
oins in relational databases, and variable elimination for solving lin
ear inequalities. Algorithms for enforcing various levels of relationa
l consistency are introduced and analyzed. We also show the usefulness
of the new definition in characterizing relationships between propert
ies of constraint networks and the level of local consistency needed t
o ensure global consistency.