Raspberry Pi NTP config with fudge factors

[Gary E. Miller]

Yo Frank!

On Thu, 5 May 2016 20:23:34 -0400
Frank Nicholas <frank at nicholasfamilycentral.com> wrote:

> > 
> > Easy with the 28 refclock.  Not sure on the 20.  DO you even see
> > the coarse time with ntpq -p?  
> 
> I don’t know enough to answer regarding “coarse time”.  Here’s my
> output (FreeBSD-10.3-RELEASE):

yeah, that's a problem with refclock #20, important data is hidden.

> > gpsd handles a lot more GPS types than #20.  So let gpsd handle the
> > GPS. gpsd passes the time it determines to ntpd using the SHM
> > (shmctl) interface.  
> 
> So the type 28 (SHM) is coming from gpsd on the same box?

Yes.  Or sometimes a PTP client.

>  I still
> like KISS (no need for other GPS type info)...

When you hide important data from the user, as #20 does, you have over
simplified.

Plus, as Eric recently said, the gpsd NMEA parser is just a lot better
coded and better tested than the ntpd one.

No matter what, the refclocks will be split into separate threads and/or
daemons for security reasons.  This is already done for gpsd, so one
less thing to do.

You are correct that full blown gpsd can be overkill, but it can be, by
compile time options, pared down to almost nothing.  So the gpsd project
will need to encourage distros to build both full blown gpsd (gpsd-max?)
and a KISS version (gpsd-lite?).

> > The long term goal is to kill off #20 as no need to maintain 2 NMEA
> > parsers.  
> 
> Understood.  I still like KISS...

Sometimes you need the complexity, like the process separation for
security.

And what the military has found is that a general purpose COTS product
will beat a special purpose product almost all the time in turns
or MTBF, training, etc.

Many more eyes already know, and audit gpsd than ntpsec.

> > time1 adjusts the PPS derived time, time2 adjusts the NMEA derived
> > time.  
> 
> So for type 20 (NMEA local) & PPS, time2 would be the correct place
> to put the fudge factor (assuming time1 (PPS) is dead nuts).

Yes, except you'll need to disconnect the PPS to see it...

Another negative of excessive application of KISS.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
	gem at rellim.com  Tel:+1 541 382 8588
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 473 bytes
Desc: OpenPGP digital signature
URL: <http://lists.ntpsec.org/pipermail/devel/attachments/20160505/0a653667/attachment.bin>