- Implementing a Client
- Parsing Strings by Using StringTokenizer
- Example: A Client to Verify E-Mail Addresses
- Example: A Network Client That Retrieves URLs
- The URL Class
- WebClient: Talking to Web Servers Interactively
- Implementing a Server
- Example: A Simple HTTP Server
- RMI: Remote Method Invocation
- Summary
This chapter is from the book
This chapter is from the book
This chapter is from the book
17.7 Implementing a Server
The server is the program that starts first and waits for incoming connections. Implementing a server consists of six basic steps:
Create a ServerSocket object.
Create a Socket object from the ServerSocket.
Create an input stream to read input from the client.
Create an output stream that can be used to send information back to the client.
Do I/O with input and output streams.
Close the Socket when done.
Each of these steps is described in more detail in the following sections. As with the client, note that most of the methods described throw an IOException, so they need to be wrapped inside a try/catch block in an actual implementation.
Create a ServerSocket object.
With a client socket, you actively go out and connect to a particular system. With a server, however, you passively sit and wait for someone to come to you. So, creation requires only a port number, not a host, as follows:
ServerSocket listenSocket = Å@ new ServerSocket(portNumber);
On Unix, if you are a nonprivileged user, this port number must be greater than 1023 (lower numbers are reserved) and should be greater than 5000 (numbers from 1024 to 5000 are more likely to already be in use). In addition, you should check /etc/services to make sure your selected port doesn't conflict with other services running on the same port number. If you try to listen on a socket that is already in use, an IOException will be thrown.
Create a Socket object from the ServerSocket.
Many servers allow multiple connections, continuing to accept connections until some termination condition is reached. The ServerSocket accept method blocks until a connection is established, then returns a normal Socket object. Here is the basic idea:
while(someCondition) {
Socket server = listenSocket.accept();
doSomethingWith(server);
}
If you want to allow multiple simultaneous connections to the socket, you should pass this socket to a separate thread to create the input/output streams. In the next section, we give an example of creating a separate thread for each connection.
Create an input stream to read input from the client.
Once you have a Socket, you can use the socket in the same way as with the client code shown in Section 17.1. The example here shows the creation of an input stream before an output stream, assuming that most servers will read data before transmitting a reply. You can switch the order of this step and the next if you send data before reading, and even omit this step if your server only transmits information.
As was also discussed in the client section, a BufferedReader is more efficient to use underneath the InputStreamReader, as follows:
BufferedReader in = new BufferedReader (new InputStreamReader(server.getInputStream()));
Java also lets you use ObjectInputStream to receive complex objects from another Java program. An ObjectInputStream connected to the network is used in exactly the same way as one connected to a file; simply use readObject and cast the result to the appropriate type. See Section 13.9 (Serializing Windows) for more details and an example. Also see Section 17.9 (RMI: Remote Method Invocation) for a high-level interface that uses serialization to let you distribute Java objects across networks.
Create an output stream that can be used to send information back to the client.
You can use a generic OutputStream if you want to send binary data. If you want to use the familiar print and println commands, create a Print_Writer. Here is an example of creating a PrintWriter:
PrintWriter out = new PrintWriter(server.getOutputStream());
In Java, you can use an ObjectOutputStream if the client is written in the Java programming language and is expecting complex Java objects.
Do I/O with input and output streams.
The BufferedReader, DataInputStream, and PrintWriter classes can be used in the same ways as discussed in the client section earlier in this chapter. BufferedReader provides read and readLine methods for reading characters or strings. DataInputStream has readByte and readFully methods for reading a single byte or a byte array. Use print and println for sending high-level data through a PrintWriter; use write to send a byte or byte array.
Close the Socket when done.
When finished, close the socket:
server.close();
This method closes the associated input and output streams but does not terminate any loop that listens for additional incoming connections.
Example: A Generic Network Server
Listing 17.17 gives a sample implementation of the approach outlined at the beginning of Section 17.7. Processing starts with the listen method, which waits until receiving a connection, then passes the socket to handleConnection to do the actual communication. Real servers might have handleConnection operate in a separate thread to allow multiple simultaneous connections, but even if not, they would override the method to provide the server with the desired behavior. The generic version of handleConnection simply reports the hostname of the system that made the connection, shows the first line of input received from the client, sends a single line to the client ("Generic Network Server"), then closes the connection.
Listing 17.17 NetworkServer.java
import java.net.*;
import java.io.*;
/** A starting point for network servers. You'll need to
* override handleConnection, but in many cases listen can
* remain unchanged. NetworkServer uses SocketUtil to simplify
* the creation of the PrintWriter and BufferedReader.
*/
public class NetworkServer {
private int port, maxConnections;
/** Build a server on specified port. It will continue to
* accept connections, passing each to handleConnection until
* an explicit exit command is sent (e.g., System.exit) or
* the maximum number of connections is reached. Specify
* 0 for maxConnections if you want the server to run
* indefinitely.
*/
public NetworkServer(int port, int maxConnections) {
setPort(port);
setMaxConnections(maxConnections);
}
/** Monitor a port for connections. Each time one is
* established, pass resulting Socket to handleConnection.
*/
public void listen() {
int i=0;
try {
ServerSocket listener = new ServerSocket(port);
Socket server;
while((i++ < maxConnections) || (maxConnections == 0)) {
server = listener.accept();
handleConnection(server);
}
} catch (IOException ioe) {
System.out.println("IOException: " + ioe);
ioe.printStackTrace();
}
}
/** This is the method that provides the behavior to the
* server, since it determines what is done with the
* resulting socket. <B>Override this method in servers
* you write.</B>
* <P>
* This generic version simply reports the host that made
* the connection, shows the first line the client sent,
* and sends a single line in response.
*/
protected void handleConnection(Socket server)
throws IOException{
BufferedReader in = SocketUtil.getReader(server);
PrintWriter out = SocketUtil.getWriter(server);
System.out.println
("Generic Network Server: got connection from " +
server.getInetAddress().getHostName() + "\n" +
"with first line '" + in.readLine() + "'");
out.println("Generic Network Server");
server.close();
}
/** Gets the max connections server will handle before
* exiting. A value of 0 indicates that server should run
* until explicitly killed.
*/
public int getMaxConnections() {
return(maxConnections);
}
/** Sets max connections. A value of 0 indicates that server
* should run indefinitely (until explicitly killed).
*/
public void setMaxConnections(int maxConnections) {
this.maxConnections = maxConnections;
}
(continued)
/** Gets port on which server is listening. */
public int getPort() {
return(port);
}
/** Sets port. <B>You can only do before "connect" is
* called.</B> That usually happens in the constructor.
*/
protected void setPort(int port) {
this.port = port;
}
}
Finally, the NetworkServerTest class provides a way to invoke the NetworkServer class on a specified port, shown in Listing 17.18.
Listing 17.18 NetworkServerTest.java
public class NetworkServerTest {
public static void main(String[] args) {
int port = 8088;
if (args.length > 0) {
port = Integer.parseInt(args[0]);
}
NetworkServer nwServer = new NetworkServer(port, 1);
nwServer.listen();
}
}
Output: Accepting a Connection from a WWW Browser
Suppose the test program in Listing 17.18 is started on port 8088 of system1.com:
system1> java NetworkServerTest
Then, a standard Web browser (Netscape in this case) on system2.com requests http://system1.com:8088/foo/:, yielding the following back on system1.com:
Generic Network Server: got connection from system2.com with first line 'GET /foo/ HTTP/1.0'
Connecting NetworkClient and NetworkServer
OK, we showed the NetworkClient and NetworkServer classes tested separately, with the client talking to a standard FTP server and the server talking to a standard Web browser. However, we can also connect them to each other. No changes in the source code are required; simply specify the appropriate hostnames and port. The test server is started on port 6001 of system1.com, then the client is started on system2.com, yielding the following results:
Time t0, system1:
system1> java NetworkServerTest 6001
Time t1, system2:
system2> java NetworkClientTest system1.com 6001
Time t2, system1:
Generic Network Server: got connection from system2.com with first line 'Generic Network Client'
Time t3, system2:
Generic Network Client: Made connection to system1.com and got 'Generic Network Server' in response