re here i was tripping upon the ntpd quality and got this idea to add message return to the sendmsg(2) system call. similar to the receive timestamps but the different -- xmit timestamps. basically there is race still in xmitting ntp packets between fetching the time and calling sendto(2) where we could be rescheduled and have a substantial error as a result.
as far as i see there is nothing that prevents control messages from being returned back (space provided) from sendmsg(2). this sounds also backwards compatible since existing callers do not provide any control message space. any thoughts? cu -- paranoic mickey (my employers have changed but, the name has remained)