On Mon, Nov 17, 2008 at 1:10 AM, Song Haibin <melodysong@xxxxxxxxxx> wrote: > Bruce, > >>-----Original Message----- >>From: p2psip-bounces@xxxxxxxx [mailto:p2psip-bounces@xxxxxxxx] On Behalf Of >>Bruce Lowekamp >>Sent: Friday, November 14, 2008 3:44 AM >>To: Das, Saumitra >>Cc: p2psip@xxxxxxxx; ietf@xxxxxxxx >>Subject: Re: [P2PSIP] P2PSIP diagnostics: PING discussion >> >>Neither XP nor Mac OS X can be relied on to have accurate time sync, >>although they in theory come with time syncing enabled by default. >>Unfortunately, you just can't rely on synchronized clocks in all but >>the most controlled of circumstances (and I think it's almost always a >>poor idea, even then). > > But one way delay (OWD) measurement is useful according to the motivations > in section 2.1 of RFC2679. It is also useful in P2P networks, e.g. when the > p2p network is used as a tunnel for data transfer. Clock synchronization is > hard for ordinary host, but that depends on what precision you want to > achieve. > Song, No dispute that it's useful, it's just that what you need to do anything useful with OWD is an order of magnitude beyond what you can expect on any p2p overlay not specifically designed for measuring it. In 2679, Section 3.5, 2nd bullet the requirement is described as having a GPS receiver on each host or maybe each host being synchronized directly with multiple gps-provided NTP servers, though they didn't test whether that was good enough. Bruce _______________________________________________ Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf