PPS and selection algorithm issues

[Jim Pennino]

On Sun, Sep 22, 2024 at 01:29:26PM -0700, Hal Murray wrote:
> 
> > refclock nmea flag1 0 baud 9600 minpoll 4 time2 0.097
> > refclock pps minpoll 4 flag2 0 time1 0.002 
> 
> Try adding a prefer to the nmea line.  THe PPS driver needs some place to 
> get the seconds.  The prefer keyword is a hack to do that.

I tried this and it worked. All the numbers are back to being under 2
microseconds where they were with legacy ntp.

*NMEA(0)         .GPS.            0 l    7   16  377   0.0000  -0.8929   0.9211
oPPS(0)          .PPS.            0 l    6   16  377   0.0000   0.0076   0.0017

I also tried going back to having 3 sources without the prefer on the
nmea line as it was with legacy ntp and got this:

*NMEA(0)         .GPS.            0 l   10   16  377   0.0000   0.3839   0.8317
xPPS(0)          .PPS.            0 l    9   16  377   0.0000   1.7825   0.9609
+192.168.0.100   .PPS.            1 u   46   64  377   0.1330   0.9937   0.7232
+192.168.0.101   .PPS.            1 u   51   64  377   0.1334   0.9648   0.6534

So it appears that prefer is absolutely required in the nmea configuration
line for the pps driver to work.

This is a big change change from legacy ntp and really needs to be
documented.

If the Linux distros all deprecate legacy ntp and switch to ntpsec,
every server on the planet with a nmea/pps reference clock is going to
break.

I would suggest the following:

In the refclock command documentation, the fact that the prefer option
is required for nmea/pps to work needs to be documented.

It probably wouldn't hurt to repeat this in the nmea and pps driver
documentation.

The documentation home page needs an entry for transitioning from legacy
ntp to ntpsec. The need for prefer needs to be here as well and some
examples of going from psuedo ip addresses to the named driver method.

I also noticed that nowhere in the documentation is the need for
setserial and ldattch documented in setting up a stratum 1 server.

Thanks for the help.