The difference between, for example, reading a local file on a single machine and performing the same read on a remote file in DCE is like the difference between reading information from a phone book yourself and dialing an operator for the same information. The remote operation requires the addition of another active entity that can be requested to perform it, since you cannot. Associated with every piece of remote data available on a network is a remote server to manage that data and make it available. The user may not see the server; even the client may be unaware of it, but it is there.
The DCE documentation often speaks of "binding to an object." In reality, clients can bind only to servers, which then may be requested to perform operations on objects that are under their management. However, it is possible for a server to put bindings into namespace entries that are named for the objects that it manages. Furthermore, these exported bindings can be tagged with object UUIDs in such a way that incoming remote calls from clients can be applied by the server to the object whose name entry the binding was read from (the details of this technique are described later in this topic). When an application uses this kind of binding model, it is reasonable to say that the client is logically bound to the object, although it is physically always bound to the server that manages the object.