To get the latest clean install of Java go to this page:
http://java.sun.com/getjava
Then click on the download icon at top of the page.
This will install correctly Java onyour machine.
If the automated upgrade does not work, you may hae a corrupted install of Java.
In that case:
- open the Windows control panel and click on Install programs/add hardware
- search Java in the list of programs, and unistall it: it will remove all corrupted registry keys
Then reinstall Java from the above link.

If this still does not work, delete the directory where Java is installed, and open regedit, goto:
HKEY_LOCAL_MACHINE
Software
JavaSoft
Java RunTime Environment
Delete the key marked "Java RunTime Environment"
Then reinstall Java from the above link.
When the Java applet displays an animated cup of coffea, you can test the performance of your new Java RunTime by visiting the Java ShowCase on the same site , notably the 3D demos. It will convince you that Java 1.4.1 is great and adds a lot of performance gains face to Java 1.3, because its memory manager has been a lot enhanced and because its internal runtime optimizer generates better native code on both Windows and Linux.

For Macintoshes, the Java runtime is on the Apple web site:
http://www.apple.com/java
For now MacOSX does not have Java 1.4 but porting it is in progress by Apple.

For Mac OS 8/9, the only version supported by Apple is Java 1.1.8 (There's no support of the Java2 API, and Apple does not want to support Java2 on Mac OS 8/9). Note: MacOS 8/9 users should really consider adding memory upgrades to 128MB if they don't want to buy the upgrade to MacOSX. This will greatly enhance the user experience by improved performance and less disk swap in the virtual memory manager, notably if you want to use the QuickTime player used by Limewire to play MP3 files.

LimeWire can run on Mac with just 32MB of memory, but some users of MacOS9 have experienced "OutOfMemory" errors when running LimeWire. This problem does not occur with MacOS8, as long as there are not too many apps running in parallel.