[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gpsd-dev] GPSD's assumptions about time
From: |
Eric S. Raymond |
Subject: |
Re: [gpsd-dev] GPSD's assumptions about time |
Date: |
Thu, 28 Nov 2013 08:30:55 -0500 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
Hal Murray <address@hidden>:
> >> There was some discussion in the ntp lists a month or two ago about a NMEA
> >> device that was broken and going to need this logic. One of them might
> make
> >> a handy test case. I'll try to track down the details if you are
> interested.
>
> > I am.
>
> It started on the time-nuts list:
> [time-nuts] Warped back to 1993
> http://www.febo.com/pipermail/time-nuts/2013-August/079197.html
> The time-nuts archive stuff often breaks threads into several chunks. If you
> want the whole thing, go back to the Thread index when you get to the end.
> Look down a bit to pick up the next chunk.
>
> Then it went over to ntp-questions:
> [ntp:questions] Start of new GPS 1024 week epoch
> http://lists.ntp.org/pipermail/questions/2013-August/035889.html
> Lots of noise here. It's a long thread.
>
> The problem has been observed in
> FURUNO GT-74, GT-77, and GT-80
> Truetime XL-DC (some, not all)
>
> Furuno had a web page:
> http://www.furuno.co.jp/en/news/notice/20130214_001.html
>
> Somewhere in there I saw a comment about using the leap second offset to get
> the right epoch. Sounds like a neat hack. That would give you 20 years from
> the time you built the leap-second table.
OK, I've read the threads. Here are my thoughts.
First: Hal, thank you for the excellent explanation of why GPSes have trouble
finding sats on a cold boot. I did not know about the role of Doppler shift!
I have incorporated this into the FAQ. You continue to be exceptionally good
at expository writing, an uncommon skill in our peer group.
(Harlan, scratch my gloomy prognostication of a few weeks ago. Hal could
probably do about as good a job of rewriting the NTP docs as I could. Not
that I'm volunteering him...)
Second, I actually looked into trying to back the era out of the leap-second
offset a couple of years back. It can't be made reliable enough. Discussion:
http://esr.ibiblio.org/?p=2869
http://esr.ibiblio.org/?p=2882
Third, I now think a brutally simple solution to the GPS-era problem
will work acceptably, and is in any case better than what we have.
I will describe it in my next post.
--
<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>
Re: [gpsd-dev] GPSD's assumptions about time, Hal Murray, 2013/11/28