The following program - due to Gilad Bracha - illustrates a loophole in the
current specification for generics when wilcards are used. This problem
and its solution are discussed in Viroli and Igarashi's papers, but we
haven't yet incorporated a solution to the problem into the specification or
its reference implementation.

//////////////////////////////////////////////////////////
import java.util.*;

class SuperOfFlaw<S>{
    S s;
    S get() { return s;}
    void put(S a) { s = a;}

    SuperOfFlaw(S a) { s = a;}
}

class Flaw<T> extends SuperOfFlaw<List<T>> {
    List<T> fetch(){return s;}

    Flaw(T t){super(new ArrayList<T>()); s.add(t);}
}

class SuperOfShowFlaw {

    SuperOfFlaw<List<?>> m(){return null;}
}


public class ShowFlaw extends SuperOfShowFlaw {

    static Flaw<Number> fn = new Flaw<Number>(new Integer(3));

    Flaw<?> m(){return fn;}


    public static void main(String[] args) {
SuperOfShowFlaw sosf = new ShowFlaw();
SuperOfFlaw<List<?>> sof = sosf.m();
List<String> ls = new ArrayList<String>();
ls.add("Smalltalk rules!");
sof.put(ls);
Number n = fn.get().get(0);
    }

}