Clients and servers

In the client-server model of computing, a number of software systems, the clients, may require general-purpose services provided by other systems, the servers. Clients and servers run concurrently, on the same machine or, more generally, on different machines connected through a network.

Typically, the clients are systems, for example systems that provide useful facilities for human users; the servers take care of operations that clients could not perform on their own (or would perform less efficiently), and of operations that are common to several clients. Example of servers include:

  • Computation servers, which take care of heavy or specialized computations not appropriate for the clients, assuming the servers' processors are more powerful.
  • File servers and database servers, providing access to persistent data that different clients may need to access.

The client-server relation is usually many-to-many: each client application, in the course of performing its duties, may successively require services from different servers; and each server may cater to different clients. This second aspect, illustrated by the following figure, is particularly important; it allows a single server to communicate with as many clients as appropriate.


Although communication occurs both ways - clients sending objects to servers and conversely - the situation in such a scheme is usually not symmetric. Whereas a client requesting a certain service will need to know beforehand what server provides that service, the server, for its part, will not in general name its clients in advance; it will simply stand ready to provide its services to whatever suitable client happens to request them.

This dissymmetry appears clearly in EiffelNet and in the examples of this manual. It means in particular that clients and servers act differently at the beginning of their client-server lifecycle. The party that initiates the communication is the client; in this operation it will identify the desired server (through a local file name for communication on the same machine, or though a machine name for network communication). The server simply makes itself ready for possible connections by starting to "listen" on the communication channels - relying for this on an EiffelNet procedure that is indeed called listen.