Re: [gpsd-dev] Clarifications needed for the time-service HOWTO

[Harlan Stenn]

"Gary E. Miller" writes:
> Yo Eric!
> 
> On Tue, 22 Oct 2013 10:08:05 -0400
> "Eric S. Raymond" <address@hidden> wrote:
> 
> > Gary E. Miller <address@hidden>:
> > > > Would you please expand this into a brief discussion of NTP
> > > > diagnostic tools for the HOWTO?
> > >=20
> > > The tools we are talking about are for snooping on other peoples
> > > chimers.  I would think that is out of scope for this document which
> > > whould be focused on how a user makes his own server a goo Stratum
> > > 1.
> >=20
> > You told me that for cross-checking purposes one needs to have at
> > least two chimers other than the local GPS configured.  I believe
> > that, it makes logical sense.
> 
> Good, but now I am leaning toward the ntp.pool.org recommendation of
> 4.
> 
> They add them to the ntp.conf file this way:
> 
> server 0.pool.ntp.org
> server 1.pool.ntp.org
> server 2.pool.ntp.org
> server 3.pool.ntp.org

The prefererred line is:

 pool <location>.pool.ntp.org iburst

> They monitor their pool and only use chimers of a certain quality.
> =46rom what I can tell they think +/-20 mSec is fine, which is a bit
> coarse for our purposes, but a start.  Maybe usefull as a minimmum
> acceptable.
> 
> > It seems to me that means someone trying to set up a Stratum 1 has
> > all the verification issues that anyone setting up an ordinary client
> > does. =20
> 
> I would say we prefer more verification than pool.ntp.org, but they have
> been at it a while longer.

How about we check with Ask and find a way to work together on this.

One of those steps would be to point folks at a good central location
for this information.

> > So some guidance to chimer auditing tools seems required.
> 
> That hhas been discussed at length on the list.  ntptrace is the
> old way, mature but deprecated.  sntp is the new way, nice andd shiny
> but moving too fast to document.  ntpq somewhere in the middle.

In what way is sntp a moving target?  That code has been stable and
pretty much unchanged for a while.

ntptrace needs an overhaul, and it and ntpq both use mode 6 packets.

I'm a bit lost - exactly what are the goals here?

> I like ntpq since it shoow you what your local ntpd is doing relative
> to the external chimers you have picked.
> 
> Also I like Beat Bolli's idea for scripting the fudge math.  That
> should be mentioned (and credited) but I'd put the details in a
> script gpsd includes to it just works.
> 
> My gut feel is that we are gonna end up writing our own tools.  :-)

If they are tools for NTP would it be better if they were in the NTP
distribution?

H