1. 38
  1. 9

    I posted this because I saw it floating around on Twitter and HN, but really didn’t understand what the problem was. I thought I’d get some feedback here.

    I assumed that GPS time was always basically TAI and didn’t do any leap second calculations, but I was wrong.

    This article lays out the issue in a way that is easier to understand.

    1. 1

      Global warming should always cause the Earth’s rotation to slow down due to shift of mass from poles to equator. Whatever is causing the speedup, I don’t think it is global warming.

      1. 3

        I was curious and googled it. I found a Forbes article which links to this research article. I don’t quite comprehend the language in the research article, but Forbes’ dumbed-down version is that as glaciers melt on and near the poles, there’s less weight on the ground, which makes the ground “rebound” and so the earth circularizes. No doubt the effect you describe also exists, mass trapped in the form of polar ice melts and moves towards the equator, but (according to my understanding of Forbes’ explanation of the paper) that effect is smaller than the counteracting effect from the rising ground.

        EDIT: Here’s a non-Forbes source, although it doesn’t contain an explanation of the mechanism: https://phys.org/news/2021-01-earth-faster.html

        They [planetary scientists] also have begun wondering if global warming might push the Earth to spin faster as the snow caps and high-altitude snows begin disappearing.

        1. 1

          In case you’re interested, Hudson Bay is one of the most well-known examples of the “rebound” effect… https://earthobservatory.nasa.gov/images/147405/rebounding-in-hudson-bay.

        2. 2

          I was also curious about this, as that matches my intuition as well.

          This article is about the shift in the axis of rotation, not the speed, but it outlines some unintuitive effects of climate change on the Earth’s distribution of mass:


          To summarize:

          • The Earth is not round; the poles are flattened, because the weight of the glaciers squishes them down. When they melt, the earth beneath them rise, and the planet becomes more round.
          • Most of the melting glaciers are in Greenland, which is 45° from the North Pole, so the mass redistribution is not as clear cut as it seems.
          • The mantle might be doing convection stuff that I cannot begin to understand. Cursory research implies this is a recent theory and maybe not totally accepted or understood. (Certainly not by me. I have no idea here.)

          Note that the article doesn’t claim any of these affect rotation speed, but they’re interesting factors for my mental model of the changing Earth. Other sources I found cite the first reason (elongating Earth) as the primary driver for the speedup of the Earth’s rotation.

          Another possible factor I found is that high-altitude snow and ice is melting, distributing mass closer to the center of the Earth (which, you know, angular momentum or something). It seems incredible to me that that would have a measurable effect – and I didn’t look for evidence that it does – but it’s another interesting thing that I would not have guessed.

          So yeah; it’s complicated. Some aspects might slow the Earth in isolation; some might speed it up. Taken together it seems like the consensus is a net speedup.

          Obligatory caveat: I probably spent more time writing this comment than actually researching this topic, so this information comes with no warranty express or implied.

          (I accidentally replied to mort’s sibling comment, which was posted about the same time; I deleted it and re-posted it here as I had not actually read it at the time I wrote this. Perils of composing in a separate editor and pasting it in.)

          1. 1

            I have to think that moving mass from the poles to the equator is a 6,000-mile journey away from the axis of rotation, which should overwhelm elevation changes and isostatic rebound. Also, the Hudson Bay area is still rebounding slowly from glaciation, maybe a few feet over ten thousand years. The rocks in Greenland haven’t had time to rebound from the last two decades of melting.

        3. 1

          I’m surprised at how often GPS threatens to break time. Would be nice if the protocols could be updated to use less human time scales like weeks, but updating satellites is a quite brittle thing.

          1. 6

            The week number was already extended from 10 bits to 13 bits. The first satellites supporting it were launched in 2005 and the new format signal has been broadcast since 2014. For capable receivers, there is no rollover until 2137 (at which point, if GPS is still flying and hasn’t been further upgraded, the 157-year ambiguity will presumably be much easier to resolve than a 19-year one).

            This is more of a “gpsd (not GPS) tried to be clever and failed” issue. They could have simply had their logic for 10-bit week numbers resolve the ambiguity in such a way that it always returns the first matching time after the build date of the copy of gpsd itself. That would have been 100% reliable for anyone who upgrades their software at least once every two decades. If that’s not good enough then maybe they could think about a ratchet. What they ended up with instead… wasn’t really smart, it just looked that way.

            1. 1

              Do note that the exact same bug would have happened in 2003 as it is not the week counter rollover that caused it but the fact that there hasn’t been a leap second for 256 weeks which is the modulus to which the week is given for the next leap second date. Who knows how many similar footguns there are, but I’d think it would be best to utilize the fact that we now have better equipment and rework the signals to be harder to misinterpret.

              1. 2

                Do note that the exact same bug would have happened in 2003

                Yes, because it’s bad code. The solution is to have less bad code. Involving leap seconds in this was a mistake.

                but I’d think it would be best to utilize the fact that we now have better equipment and rework the signals to be harder to misinterpret.

                Again, already done. Not in the exact way you’re asking for, but in a way that works just fine in the real world. This is gpsd attempting to do its best with receivers relying on the old format (and fumbling it).

                1. 1

                  A better solution is to make it harder to write bad code. GPS handling code will be written yet again another thousand times. Do we need to keep the possibility of them having the same exact bug yet again?

                  But I do have to agree on the fact that it has been fixed. The week on which the rollover will happen is now specified with the same modulo as the week so such bugs should hopefully no longer occur with new receivers.

                  1. 1

                    The week on which the rollover will happen is now specified with the same modulo as the week

                    This statement is nonsense.

            2. 1

              It’s not just satellites, it’s every receiver deployed in the last 40 years, many in neglected but safety- or life-critical use.

              1. 1

                It’s possible to add extra signals, it has already been done. Having one with 40 bit seconds counter should definitely be possible without interfering with existing signals, and it would decrease the fragility of GPS time tracking massively.

                1. 1

                  How exactly does that improve the situation of the receivers existing in the wild?

                  1. 1

                    It improves the situation of the receivers that will exist in the wild. This exact error happened 17 years ago to some Motorola receivers. The frequency of such occurrences is rare enough for information to be common knowledge even amongst those who implement GPS, but with enough impact that letting such errors repeat is asking for trouble.