“…But it seems to us that in our approach in order to have a C k smooth barrier function W k one always has to compensate with a growth of W (·, ζ) of asymptotic order k 2 ν. But compare the preprint of Thilliez [22] where a different method is given.…”