Many biological processes rely on the interaction of proteins with multiple DNA sites separated by thousands of base pairs. These long-range communication events can be driven by both the thermal motions of proteins and DNA, and directional protein motions that are rectified by ATP hydrolysis. The present review describes conflicting experiments that have sought to explain how the ATP-dependent Type III restriction-modification enzymes can cut DNA with two sites in an inverted repeat, but not DNA with two sites in direct repeat. We suggest that an ATPase activity may not automatically indicate a DNA translocase, but can alternatively indicate a molecular switch that triggers communication by thermally driven DNA sliding. The generality of this mechanism to other ATP-dependent communication processes such as mismatch repair is also discussed.