> i think Jeff is talking about the buffering syslog() messages before they get > sent to the socket (and thus before the daemon can read it). so the logic > cant be in syslogd as that's on the other side of the socket. I understood the suggestion, and recommended against it.