Once an application creates an AWT component such as a Frame, then
even if all AWT components go away, the AWT threads still exist and
run.  This means that even if a program exits its main method, the
program will continue to run until some bit of code calls System.exit.

Below is a test case.  The following application doesn't exit unless
you uncomment the System.exit call.

import java.awt.*;
public class Test
{
    public static void main(String argv[])
    {
	Frame f = new Frame();
	f.show();
	f.hide();
	f = null;
	//System.exit(0);
    }
}

N/A

We've found several applications that unfortunately rely on that behavior, and therefore can't correct this problem because of the need for backwards compatibility.

There is a statement in 100% Pure Java documentation that you can't use
System.exit()
in order your program to be 100% pure. How to end it?

This seems to be a confirmed bug in Bug numbers 4072987 and 4038690.  The
workaround is to call System.exit(0) but it's far from ideal.

Can't the applications relying on this behaviour just
install another thread that just keeps them going even 
after the last AWT Component was disposed?
This to me appears to be a far more sensible solution than
the one currently preferred. For normal everyday usage
of AWT it is only natural to assume that the program
quit "whenever the last frame is closed".

Making old bugs ito new features? Isn't that what Microsoft keeps getting
accused of? This is just silly and dissapointing! 
This misfeature makes it REALLY difficult to determine when a complex
application using multiple windows should quit. 

even this happend while run the RMI server class also.

I would just like to say that it was a BAD call telling
people to use System.exit() to work around an obvious JVM
resource leak. And now that client side applications are
becomming more wide spread, and the need for shared jvms is
becomming more and more necessary, applications calling
System.exit() will now effectively destroy each other. 
Thanks for the forward-thinking solution.

-Chris

I have to agree with knollc.
The JVMs are going to have to start supporting multiple 
applications and this makes that impossible.

I also agree with sz9.  The programs have an easy way out.
Just create a thread that doesn't terminate.

Please don't hold up future progress, because of such a 
glaring bug. 

I believe this was fixed in 1.4.  See 4030718.