ver_p depends on the amount of data we have read ahead, so it is not
suitable to get predictable results for ESI testing