Re: [gpsd-dev] Meaning of negative PPS offset?

[Greg Troxel]

  3. When doing 2, gpsd looks at the last timestamp received in the
  GPS datastream, assumes it went with the *previous* reporting cycle,
  and increments the second part by one second.  This is assumed to be the
  base clock second for the *current* reporting cycle.

I think this is an unwarranted assumption and the root of the issue.

Really, the GPS receiver has its own embedded computer, which is doing
who knows what.  Fixes can be computed at any time, and projected.   So
someone may have thought it clever to compute a fix just before the
top-of-second, and to project forward to the second, and emit it "on
time".

The notion of "reporting cycle" is basically an assumption about the
firmware in a receiver.   Probably for some value of "cycle" it is a
reasonable assumption.

So in general I think each receiver needs a characaterization of its
typical serial delay, and pps timestamps have to be disambiguated with
respect to that.

It might be that assuming serial is 100 ms early to 900 ms late is good
enough for all actually built receivers.