The constraint was applied prior to decoding, and rejected anything containing the '%' character, which would be used for anything with non-ASCII unicode characters.