Richard Laager said:
> I believe you're looking for "fake-hwclock". It periodically saves the time
> to a file (allegedly*  /etc/fake-hwclock.data) and restores it on boot. 

Thanks.

I discovered fake-hwclock via Google but it wasn't on my system and the 
discussion I was looking at was very old so I assumed it had been replaced or 
something.  Looks like Ubuntu just didn't include it in their img file that I 
used.

I installed it and things are happy.

I think this is a good-enough solution.

Basically, we have to document that the clock has to be close-enough and what 
close-enough means.

Cold-stanbys sitting on the shelf for many years won't work.
PCs with dead CMOS/TOY clock battries won't work.

(Let's Encrypt certificates are only valid for 90 days.)

IoT devices on a store shelf for a few months might not work.

Working Raspberry PIs that sit on the shelf for long enough may not work.

You can fix the stanby/store shelf problem by carefully setting up certificates 
with a long lifetime.

------

> I still think we need a more comprehensive approach to this bootstrapping
> problem. The problem is, I don't have the time to write it. But I gave my
> thoughts before: https://lists.ntpsec.org/pipermail/devel/2019-February/
> 007576.html 

That looks reasonable, but complicated.

I'm not planning to work on anything like that.

------

>> That could backfire if, somehow, the system time got set into the future.
>I had that happen once. It might have been due to a GPS rollover.

GPS rollover seems more likely to go the other way -- that is step back by 20 
years.

It would go forwards if it was broken, you used fudge to fix it, and then the 
software got fixed and you installed the new software without removing the 
fudge.



-- 
These are my opinions.  I hate spam.



_______________________________________________
devel mailing list
devel@ntpsec.org
https://lists.ntpsec.org/mailman/listinfo/devel

Reply via email to