We have been thinking also about the possibility of getting a direct proof of Invariant Approximation Property for amenable groups that uses some averaging procedure and does not use von Neumann algebras as in John Roe ...