[time-nuts] NTP stratum 0

Magnus Danielson cfmd at bredband.net
Sat Jan 14 13:29:45 EST 2006


From: "M. Warner Losh" <imp at bsdimp.com>
Subject: Re: [time-nuts] NTP stratum 0
Date: Sat, 14 Jan 2006 10:53:40 -0700 (MST)
Message-ID: <20060114.105340.89662706.imp at bsdimp.com>

> In message: <20060114164520.GA25156 at feynman>
>             Neal McBurnett <neal at mcburnett.org> writes:
> : On Sat, Jan 14, 2006 at 05:25:04PM +0100, TwoSpades wrote:
> : > I would like to know where could I find the FORMAL specification of the NTP
> : > stratum 0.
> : > I often read that it can be a GPS receiver, but I think that there must be a
> : > formal definition
> : > of the characteristics of that device.
> 
> Well, it depends on what you mean by formal.  The 'rough and ready'
> definition for stratum 1 time servers is that they are synchronized to
> a time source better than they would get from another ntp server.  If
> your system time can be synchronized to within a few tens of
> microseconds with the receiver you have, then you are definitely
> stratum 1 material.  If you relax the requirement to be only as good
> as you need to be to serve time over the internet, then I'd say that
> requirement drops to about 1ms.

Indeed. But formally, also an undiciplined XO can be put as a Stratum 1, so one
should not assume that you get even near UTC tracking. The Stratum isn't really
a clock quality measure, it is rather a very rougth and quite possibly even a
missleading measure of additional wander from the primary source, if we are
compare same source. Even the NTPv4 spec seems to be missing the necessary
support to handle things. Will check the details.

> In general, you want the server to be traceable back to the official
> time, as produced by some time authority.  For a simple ntp server,
> that tracking doesn't have to be too precise, from a time geeks point
> of view.  If you are within about a millisecond, then no one will
> notice if the time exchange is over the internet...

Indeed. But no self-proclaimed time nut would dream of providing an NTP which
ain't overspeced in at least a couple of aspects. :-)

Personally I am working on a solution, but too little time is the main stumble
block.

Cheers,
Magnus



More information about the time-nuts mailing list