My line of thought is that a Russellian theory of meaning is kind of baked into the idea of a classification problem, and that class membership of real world things doesn't always behave as discretely as Russell would have liked.