> As a user, I would *not* expect that there's just one fooCount field for > all Foo<T> instances. Except that's how generics have worked in Java since 2004. Do you propose that existing generic code with statics should change how it behaves? Or are you suggesting that the existing behavior is also counterintuitive?