13.5. Using Native Methods
Despite Sun's push for 100% Pure Java, native code still has its place. You need native code to do things that Java (and external programs launched by Java) cannot do: locking files, accessing user IDs, accessing shared memory, sending faxes, and so on. Native code is also useful when accessing legacy data through non-Java gateways. Last, in situations where every last bit of performance is vital, native code libraries can give a servlet a big boost.
Native code, however, should not be used except when absolutely necessary, since if the native code run by a servlet goes south, the entire server goes down with it! The security protections in Java can't protect the server from native code crashes. For this reason, it's wise not to use the native JDBC-ODBC bridge from a servlet because many ODBC drivers seem to have problems with multithreaded access. Native code also limits the platform independence of a servlet. While this may not matter for custom-built servlets tied to a particular server, it's something to remember.
How a servlet accesses native methods depends on the web server and JVM in which it's running. To take a risk and speak in broad generalities, let us say that you can pretty much expect your web server and JVM to support the standard Java Native Interface (JNI). Using JNI is fairly involved, and even a basic introduction extends beyond the scope of this chapter. For a tutorial on JNI, see the upcoming Java Native Methods, by Alligator Descartes (O'Reilly).
When using JNI with servlets, remember these things:
Copyright © 2001 O'Reilly & Associates. All rights reserved.