Modularization is an important architectural principle underlying many types of complex systems. It tends to tame the complexity of systems, to facilitate their management, and to enhance their flexibility with respect to evolution. In software, modularization has been practiced and studied thoroughly in local, i.e. non-distributed systems. But very little attention has been paid so far to modularization in distributed systems. This is, in part, because distributed systems are inherently modularized, in the sense that the internals of each component of such a system is inaccessible to other components, thus satisfying the Parnas hiding principle. It is, however, the thesis of this paper that there is much to be gained by being able to treat groups of distributed components as modules, called here distributed modules. And that besides the conventional hiding principle, distributed modularization should provide additional capabilities, which rarely, if ever, figure in conventional modularized systems. These capabilities include, but are not limited to: the ability to impose constraints on which kind of messages can be sent from a given distributed-module to its outside; and the ability to create AOP-like crosscutting modules. This paper introduces a model of modular distributed system, orMDS, which satisfies such capabilities, and which is implemented via the LGI middleware.