Stability for the mailing problem