Is there a security model that's both highly secure, and foolproof regardless of the mental faculties of potentially billions of diverse users? I think the answer is, "Obviously not," so the real question is whether or not the necessary compromises made here represent acceptable measures.