In a Service-Oriented Architecture (SOA), a system is viewed as a collection of independent units (services) that interact with one another through message exchanges. Established languages such as the Web Services Description Language and the Business Process Execution Language allow developers to capture the interactions in which an individual service can engage, both from a structural and from a behavioral perspective. However, in large serviceoriented systems, stakeholders may require a global picture of the way services interact with each other, rather than multiple small pictures focusing on individual services. Such "global models" are especially useful when a set of services interact in such a way that none of them sees all messages being exchanged, yet interactions taking place between some services affect the way other services interact. An issue that arises when dealing with global models of service interactions is that these models may capture behavioral constraints that can not be enforced locally. In other words, some global models may not be translatable into a collection of local models such that the sum of the local models equals the original global model. Starting from a previously proposed language for global modeling of service interactions, this paper defines an algorithm for determining if a global model is locally enforceable and an algorithm for generating local models from global ones.