yes if we supported computed permissions on the LHS of an arrow that would be an alternative, but that's a pretty big deviation from the zanzibar paper and it will tank performance of distributing subproblems