Enterprise-level applications use a lot of different directory services-lookup services that locate resources associated with a particular name. When you use RMI, for example, you locate objects with a directory service called the RMI Registry. When you use CORBA, you use the COS Naming facility (CORBA's naming service) to locate objects. When you convert a hostname to an IP address, you usually use a directory service called DNS (Domain Name Service). There are also general directory services that use protocols, such as X.500 (the CCITT directory standard) and LDAP (Lightweight Directory Access Protocol). These directory services can hold many kinds of data.
Although most people tend to use the terms "naming service" and "directory service" interchangeably, there is a difference. A naming service associates a single name with a specific resource. A directory service associates a name with a set of attributes and resources. When you search a naming service, you can only search for a specific name. When you search a directory, you can search for items matching a specific set of attributes.
One of the interesting things about all these types of naming and directory services is that they generally perform the same task-mapping a name to some set of attributes or objects. Of course, not all directory services are created equally. Some of them have a flat namespace, whereas others offer a tree structure for the names. Some of them allow you to store specific types of objects, whereas others allow you to store almost any kind of object.
The Java Naming and Directory Interface (JNDI) draws a distinction between naming services and directory services. A naming service maps a name to an object. The RMI Registry and the CORBA Naming Service are both examples of naming services. You can only store an RMI object in the RMI Registry and you can only store a CORBA object in the CORBA Naming Service. A directory service also stores objects, but these objects can have associated attributes that the directory service recognizes. You can search a directory using the item attributes. For example, you can search an LDAP directory for everyone in a specific department or everyone named Smith.
JNDI provides a uniform way to access naming and directory services. It supports flat namespaces as well as tree namespaces, and it allows you to store many different types of objects. The beauty of JNDI lies it its simplicity and uniformity. After you know the basic JNDI API calls, you can read data out of any kind of directory as long as there is a JNDI service provider for that directory.
You have already encountered JNDI in several earlier chapters. You use JNDI to locate Enterprise JavaBeans and JDBC connection pools from within your EJB container. You might have implemented simple lookup schemes before in your applications; that is, you create a class with static lookup methods or store a Hashtable in a static field somewhere. You might choose to use JNDI to replace these kinds of local storage mechanisms, although you might need to write your own service provider.
JNDI is also extremely useful in the area of configuration. If many applications use common configuration data, you might consider storing the data in a directory service, such as LDAP, instead of in a file or database. LDAP is especially good if the configuration information is hierarchical-that is, if it is more like a tree structure than a flat list of values.
One of the hidden benefits of directory services is the fact that there are a lot of directory service browsers and editors-especially for LDAP. You can view the contents of the directory and edit them using an off-the-shelf tool. That saves you from having to write a custom configuration editor.
The Context class is the core of the JNDI API. You use it to perform any lookup and to add any new name-value associations. When you use JNDI, you typically create an InitialContext object first:
Context ctx = new InitialContext();
The InitialContext constructor looks for a system property called java.naming.factory. initial that contains the name of the class that creates the InitialContext. Sometimes, you must supply this value yourself. Some EJB containers, like the one that comes with Sun's J2EE SDK, already have this property set.
JDK 1.3 comes with three built-in service providers: RMI, CORBA, and LDAP. The classnames for the different initial context factories are
com.sun.jndi.rmi.registry.RegistryContextFactory com.sun.jndi.cosnaming.CNCtxFactory com.sun.jndi.ldap.LdapCtxFactory
Don't worry about setting defining the class for the initial context factory unless you get an error telling you there's no initial context factory.
When you run your program, you can specify the initial context factory on the command-line using the -D option:
java -Djava.naming.factory.initial=com.sun.jndi.ldap.LdapCtxFactory usingj2ee.naming.JNDIDemo
You can also specify the initial context factory in a Hashtable that you can pass to the InitialContext constructor:
Hashtable props = new Hashtable (); props.put(Context.INITIAL_CONTEXT_FACTORY, "com.sun.jndi.ldap.LdapCtxFactory"); Context ctx = new InitialContext(props);
Bear in mind that if you specify the initial context factory using a Hashtable object, you might be limiting the portability of your classes. For example, most WebLogic examples tell you to create the InitialContext this way:
Hashtable props = new Hashtable(); props.put(Context.INITIAL_CONTEXT_FACTORY, "weblogic.jndi.WLInitialContextFactory"); props.put(Context.PROVIDER_URL, "t3://localhost:7001"); Context = new InitialContext(props);
The problem here is that if you want to run your code with another application server, you'll have to recompile your code with a new set of properties. It's better to set these items on the command line:
java Djava.naming.factory.initial=weblogic.jndi.WLInitialContextFactory -Djava.naming.provider.url=t3://localhost:7001 MyTestClient
Rather than specifying the initial factory on the command line, you can put these associations in a file called jndi.properties, which can be located somewhere in your classpath.
When you develop Enterprise Java Beans, you can usually count on the environment being set up properly ahead of time, so you normally don't need to initialize any properties or set any system properties. When you run your client programs to test the EJBs, however, you usually need to specify an initial context factory.
Although most people use the InitialContext object as their first entry point into JNDI, there is an alternative. You can use the javax.naming.spi.NamingManager class to create a service-specific context for you based on a URL prefix. A fully qualified JNDI name is of the form service://itemname, where service is a name such as iiop, rmi, ldap, and so on, and itemname is the name of the item in that service. The NamingManager class lets you create a Context object based on the service name. For example, to create an LDAP Context object, you can call:
Context ctx = NamingManager.getURLContext("ldap", null);
One thing to keep in mind when you use this technique is that the Context you get back usually doesn't understand names for other services. For example, if you create an initial context that is a CORBA naming service, you can still do an LDAP lookup like this:
Object ob = context.lookup("ldap://localhost/dc=wutka,dc=com");
The InitialContext object knows how to resolve references that use other kinds of services. If you try this with a context returned by getURLContext, however, you'll get an error telling you that the name isn't valid for the context you are using.
Okay, now that you have a Context object, you can use the lookup method to locate an object. For example, when you locate an EJB, you usually make a call like this:
Object personHomeRef = context.lookup( "java:comp/env/ejb/Person");
Don't forget, if you need to cast the result from context.lookup to a specific Remote or Home interface type, you must use PortableRemoteObject.narrow.
The java service is available only within an EJB container, and it acts as a local directory service for other objects within the same EJB environment.
To create a new name-value association, use the bind method:
If the object already exists in the directory, bind throws a NameAlreadyBoundException. The rebind method does the same thing as bind except that it doesn't care whether the object already exists:
rebind doesn't throw an exception if the object doesn't exist; that is, you can use rebind to create a new association as well as to overwrite an old one.
To remove an association, call unbind:
To rename an association, call rename:
You can close the InitialContext by calling the close method:
Because the context uses resources in the naming or directory service, you should close the context when you are done with it.
Make sure each EJB client creates its own InitialContext, especially if you are using EJB security credentials. The credentials are tied to the InitialContext, and if you aren't careful, one client may be using another client's credentials. Normally this isn't a problem if the clients are running as separate processes. If you're writing a Web application, however, on a server that acts as multiple clients, you must be careful to keep the contexts separated.