Can you describe what is this slightly different approach and why it should work on all models?