Asynchronous Sockets

Summary

This chapter discusses how to use network functions in the Windows graphically oriented programs—an environment that is different from the traditional Unix programming environment because it supports an event programming model. Event programming puts control of program execution in the hands of the customer. The program does not follow a normal program flow. Instead, it responds to events that occur in the course of the customer’s interaction with the program (clicking buttons, double-clicking list items, selecting menu items, and so on). Program control jumps around from method to method, reacting to events generated both by the customer and by the network.

The .NET network classes include asynchronous Socket methods that can interact within a Windows event program. These asynchronous functions do not block program execution waiting for network events. Instead, the asynchronous methods register a method from the AsyncCallback class that completes the network function when it is ready for processing. For instance, instead of a function’s blocking on a Receive() method, a BeginReceive() method registers an AsyncCallback method to run when data is available on the socket. When the data is available, the registered method is called, and the EndReceive() method reads it.

The .NET network library also supports the traditional Unix method of using non-blocking sockets. The Socket method Poll() allows a programmer to check a socket for network availability—either the availability of data if it is ready for writing, or whether an error has occurred. If the socket is not ready, the Poll() method indicates this, and the program can avoid the network function. Most programmers use a loop to check the Poll() method on a regular basis while performing other functions in between.

Another traditional non-blocking socket method is the Socket method Select(), which monitors multiple Socket objects for network availability. Similar to the Poll() method, Select() checks for the availability of data if the socket is ready for writing, or whether any errors have occurred. The ArrayList of Socket objects sent to the Select() method is monitored for a predetermined amount of time. If an event occurs on one or more of the sockets, then Select()returns with the ArrayList modified to contain only the sockets that are ready for the network action. If the time limit expires before any of the sockets are ready, the ArrayList is returned empty.

The Select() method is often used in server applications where a large number of client sockets must be monitored for activity. Each of the active sockets is placed in an ArrayList object and monitored by the Select() method for incoming data. When incoming data is detected on a group of sockets, they are passed to a method to read the data and act accordingly. As each client disconnects from the server, its Socket object is dropped from the ArrayList and no longer monitored.

The asynchronous Socket methods do all their work using threads running in background from the main application. The next chapter, "Using Threads," tells you how you can manually use threads in your own network programs to perform network functions separately from the main application