Package org.jcsp.net.cns

Class CNS

  • All Implemented Interfaces:
    CSProcess, Service
    public class CNS
extends Object
implements Service, CSProcess

    This class is the Channel Name Server's main server process class.

    This class should only be instantiated at Nodes wishing to run a server process. Although this class does not need to be used by clients wishing to interact with a server, it does provide some convenient static methods for client code to use. There are static versions of the methods provided in CNSService and there are also static factory methods for constructing CNS registered channel objects.

    Server Installation

    Channel Name Servers may be run either on a dedicated Node or else on the same Node as one of the user Nodes. The former approach is recommended for most sitations but for smaller scale use, the latter approach may suffice. The service implements the org.jcsp.net.Service interface and can be installed in the same way as any other other service by using the service manager (ServiceManager). Alternatively, the service provides install methods which handle the service manager interaction directly.

    The following example initializes a Node and installs a Channel Name Server. It then proceeds to install a CNS client service and creates and resolves a channel. The example does not proceed to do anything else but could be used as the framework for an application wishing to host its own Channel Name Server. 

     import org.jcsp.lang.*;
 import org.jcsp.net.*;
 import org.jcsp.net.cns.*;
 import org.jcsp.net.tcpip.*;

 import java.io.IOException;

 public class CNSInSameJVM implements CSProcess {

   //main method for running example
   public static void main(String[] args) {
     CNSInSameJVM proc = new CNSInSameJVM();
     proc.run();
   }

   public void run() {
     NodeKey key = null;
     NodeID localNodeID = null;
     try {
       //Initialize a Node that does not have a CNS client
       key = Node.getInstance().init(new XMLNodeFactory("nocns.xml"));
       localNodeID = Node.getInstance().getNodeID();

       //Initialize the CNS Server Process
       CNS.install(key);

       //Dedicated server code could stop here

       //Initialize the CNS client
       //use the first address of the local Node as CNS address
       NodeAddressID cnsAddress = localNodeID.getAddresses()[0];
       CNSService.install(key, cnsAddress);

       //creating Channel named "in"
       NetChannelInput in = CNS.createNet2One("in");
       //resolve the channel
       NetChannelOutput out = CNS.createOne2Net("in");

       //could now use these channels for something!!
       //but this is only a test so will terminate


     } catch (NodeInitFailedException e) {
       e.printStackTrace();
     } catch (IOException e) {
       e.printStackTrace();
     }
     Node.info.log(this, "Done.");
   }

 }

    This is the contents of the nocns.xml file:

    <JCSP-CONFIG>
    <PROTOCOLS>
    <PROTOCOL id="TCPIP" name="TCP/IP" idclass="org.jcsp.net.tcpip.TCPIPProtocolID">
    </PROTOCOL>
    </PROTOCOLS>
    <ADDRESSES>
    <ADDRESS protocolid="TCPIP" value="127.0.0.1:7896" unique="FALSE">
    </ADDRESS>
    </ADDRESSES>
    </JCSP-CONFIG>

    The above code can be copied into a file named CNSInSameJVM.java and compiled an run in same directory as the nocns.xml file.

    Channel Factory Methods

    In order to construct a ChannelInput object which can be resolved by other users of a channel name server, a client simply needs to to do this: 

       NetChannelInput in = CNS.createNet2One("Fred");

    Another process using the same channel name server can create a ChannelOutput that will send objects to this channel by do this: 

       NetChannelOutput out = CNS.createOne2Net("Fred");

    When these factory methods are called, various resources are used within the JCSP infrastructure. A channel name will be registered and held in the channel name server. These resources are taken for the duration of the JCSP Node's runtime unless the user manually frees the resources. When channel ends are constructed with these factory methods, the destroyChannelEnd(***) methods can be used to free all resources. It is only really necessary to do this if channels are being created for short term use within a long-lived Node.

    This is an example "Hello World" program which contains two inner classes with main methods, each of which can be run in separate JVMs. 

      import org.jcsp.lang.*;
  import org.jcsp.net.*;
  import org.jcsp.net.cns.*;

  public class TestCNS {

    public static class Rx {
      public static void main(String[] args) {
          try {
            Node.getInstance().init();
            NetChannelInput in = CNS.createNet2One("rx.in");
            System.out.println(in.read());
            CNS.destroyChannelEnd(in);
          } catch (NodeInitFailedException e) {
            e.printStackTrace();
          }
      }
    }

    public static class Tx {
      public static void main(String[] args) {
        try {
          Node.getInstance().init();
          NetChannelOutput out = CNS.createOne2Net("rx.in");
          out.write("Hello World");
          CNS.destroyChannelEnd(out);
        } catch (NodeInitFailedException e) {
          e.printStackTrace();
        }
      }
    }
  }

    This code can be compiled and then the following run at two command prompts:

    java TestCNS$Rx

    java TestCNS$Tx

    The programs will connect to a default channel name server. The Rx program will create a NetChannelInput and wait for a message on the channel. Once it has received the message, it prints it, destroys its channel and then terminates. The Tx program creates a NetChannelOutput that will send to the Rx program's input channel. It sends a "Hello World" message. Once this has been accepted by the Rx process, it destoys its output channel and terminates.

    CNS Client Methods

    The following code functions the same as the above code but does not use the CNS class' factory methods. The code uses the CNS client methods and manually registers and resolves the channel with the channel name server. 

      import org.jcsp.lang.*;
  import org.jcsp.net.*;
  import org.jcsp.net.cns.*;

  public class TestCNS {

    public static class Rx {
      public static void main(String[] args) {
          try {
            Node.getInstance().init();
          NetChannelInput in = NetChannelEnd.createNet2One();
          ChannelNameKey key = CNS.register(in, "rx.in");
          System.out.println(in.read());
          CNS.deregisterChannelName("rx.in", null, key);
          } catch (NodeInitFailedException e) {
            e.printStackTrace();
          }
      }
    }

    public static class Tx {
      public static void main(String[] args) {
        try {
          Node.getInstance().init();
          NetChannelLocation loc = CNS.resolve("rx.in");
          NetChannelOutput out = NetChannelEnd.createOne2Net(loc);
          out.write("Hello World");
          out.destroyWriter();
        } catch (NodeInitFailedException e) {
          e.printStackTrace();
        }
      }
    }
  }

    The CNS client methods provide the programmer with greater control over how the Channel Name Server is used. Interaction with the server need not be performed at the same time as construction of the channel. A channel can be registered with a Channel Name Server at any time after construction of its input end. A channel can be resolved at any time before construction of an output end. This allows one process to resolve a name into a NetChannelLocation object and then pass this object on for another process to use in constructing the channel. The Channel Name Server will allow a channel to be registered multiple times with different names and/or in different name spaces. Channel implementations that make direct use of the Channel Name Server may forbid this, so the behaviour of channel implemenations should be checked before this is carried out

    Author:
    Quickstone Technologies Limited
    See Also:
    CNSService, ServiceManager, Node