Reduce our reliance on NTP servers!

[stormcrow]

In the 90s, you could set a PC to the correct time and after a year, it still was very close to the correct time. Nowadays it'll run away very quickly.

This was never actually true. If you were lucky enough to have an accurate timer this was true, but the vast majority of PCs did not keep anywhere near accurate time.

As for how bad could it be if your clock is off by 10-20 mins... that's only asked by someone that's never had to deal with logging... ever. Being off by that much time in your logs makes them useless for nearly any practical or legal purposes. Try to assert "well my logs show" when the other side can immediately prove your logs are time inaccurate will destroy any case you might have even if it's to your own bosses let alone in court - this matters even if it's your personal computer because you will never know your computer's evidentiary value until it's too late to fix any deficits. Utilizing proper time stamps and logging benefit far exceeds any trivial greenwashing you can do for utilizing the public NTP service infrastructure.

[stormcrow]

Pseudoproblem. If you truly feel that you want to be "greener" and not check time through NTP, you can get atomic clock time feed from GPS satellites in the sky.
All you need is a USB GPS peripheral device connected to your PC and bit of sw setup.

Not just any USB GPS receiver will work for this. You need a receiver that also can measure and respond with its own signal latency responses otherwise the jitter between poll and response accumulates over time resulting in inaccurate clock drifts. The gpsd team has a better explanation: https://gpsd.io/gpsd-time-service-howto.html

[oiaohm]

Not just any USB GPS receiver will work for this. You need a receiver that also can measure and respond with its own signal latency responses otherwise the jitter between poll and response accumulates over time resulting in inaccurate clock drifts.

This is true. But different cases I have used more generic USB GPS receivers. The clock drift is way less than a raspberry pi3/4 that every time it restart it restores to the last time it was at when disconnected from the network. This is days/months of drift at times. Generic USB GPS receivers at least keeps this under 10 mins of drift way from correct time and remember +- 24 hours with most websites is good enough and picky ones +-30 mins that the connection will work. Yes there are a lot of data logging cases where +- 10 mins does not matter.

Yes this is are we after a perfect solution or a good enough solution. There are a lot of cases where generic USB GPS is good enough and better than the real-time clock on the motherboard if it exists. There are a lot of compact solutions that don't have a real-time clock.

I don't dispute what gpsd wrote if you are wanting really good time values you need todo what they describe. There is a question how good do we need.

Its one of the things with OS clock interfaces they don't include option to say the current clock configuration has this expected amount of error. Kind of would be useful so that particular applications could alter user to say for this application your clock error is possible too high do you want to continue anyhow or will you correct you clock system first.

[jasonjulius1122]

Hello! hope you guys are well, while minor clock discrepancies may not affect casual web browsing, certain applications and industries require precise time synchronization. Finding a balance between environmental benefits and functional requirements is crucial. Implementing local NTP servers or syncing less frequently could be options to explore.