[ntp:questions] Number of Stratum 1 & Stratum 2 Peers
unruh at invalid.ca
Sat Dec 13 02:26:41 UTC 2014
On 2014-12-13, Harlan Stenn <stenn at ntp.org> wrote:
> William Unruh writes:
>> On 2014-12-12, Harlan Stenn <stenn at ntp.org> wrote:
>> > Mike,
>> > I think you are seeing the correct and expected behavior.
>> > The root cause here is that the majority of the upstream servers are
>> > *incorrectly* not advertising the leap second.
>> > There have been problems before where a misconfigured server has
>> > incorrectly advertised a non-existent leap-second, and in cases where
>> > folks had an adequate number of correctly configured servers, this
>> > mistake was properly ignored.
>> > I have not been closely following this thread, so I might be missing
>> > something.
>> > It's pretty easy to download and install a leapsecond file, and ntpd
>> > will pay attention to that...
>> > Or am I missing something?
>> Yes, That it was an example of a case in which the correct time server
>> could be declared a falseticker. The answer you give "never use sources
>> which might not deliver the correct time" does not obviate the point.
>> The concept of "falseticker" and of "bad time" are not the same thing.
> Nobody is saying "never use sources which might not deliver correct time."
> Aside from that:
> - what is the effective difference between a falseticker and a server
> that provides "bad time"?
> - Is this difference really significant?
> - If so, how can ntpd determine the difference between a falseticker and
> a server that provides "bad time"?
It cannot easily. All it can do is decide that something is a falseticker. But
noone should think, or say, that that means that the falseticker is a
bad time source. It may be the best that ntpd can do. That does not mean
that problems cannot arise precisely because of that.
One could for example simply use all the time sources and use them all.
That has advantages and (as you will be the first to point out)
Or you could decide that your own clock is a reasonable flywheel, and
test the outside clocks over time against your won. If you find you own,
while it freewheels, roughly tracks one of the timesources, you have
additional evidence that the one may well be ticking at the right rate.
Yes, this both requires long term memory, and can certainly at times
lead you astray. I am sure I could think of other possibilities as well.
> - what real benefits does one get from knowing the difference?
Clarity in what one is talking about. Care in setting up one's system.
knowledge of what the problems can be.
Since the purpose of ntpd is to have your computer track the correct
time ( not the majority time) as closely as possible, realising that
there may be difference makes you think differently while developing a
system. What are the costs and benefits of the various possible ways of
trying to disentable the correct time from the various reports delivered
by the servers.
Maybe majority vote is the best way, but it is certainly not the only
way or even the ideal way in all circumstances.
I for one would trust time from a gps pps a lot lot more than that from
even 3 other servers. And yes, I could be fooled. Just as you might
trust advice from David Mills much more than any 5 other posters here (
and would probably be very resistant to a "majority vote" determination
that David is a "falseticker") but that does not give a guarentee that
you could not be led astray sometime.
> I'm sure a few more questions can come up, but I'm more interested in
> seeing if there might be something useful going on here, other than
> coming up with more issues that NTF's nascent Certification and
> Compliance programs should watch for.
More information about the questions