It took Facebook/Meta only like ~20 years to figure out they shouldn't rely on time to keep things in sync? /s
If you delve in distributed systems, I would be surprised if you never had the case of "hey 1/k of our computers think they are 1-2 hours behind in time. (because of NTP issues)"
You could simply be working on a single computer and have to deal with time issues. (E.g. Local clock resetting to a prior date and system creating files timestamped earlier than older files.)
I think the leap second is the most minor, not even worth mentioning problem. It is actually to our benefit, reminding us that (perceived) UTC is not a monotonic measure and computer time altogether is not. Computers nowadays adjust their time ~constantly.
That they have to smear their leap second throughout multiple hours, and that computers having the wrong time leads to an outage (for a less than a second difference) troubles me honestly. What you are telling me is that next leap second we should all pray for Meta's NTP servers not going down???
Also: (if you are an fb engineer) What do you do with negative leap seconds? Speed things up...? (https://www.timeanddate.com/time/negative-leap-second.html) The expectation is for a negative leap second coming up -- as the post does mention.
This is an education problem if you expect that time intervals have to be strictly positive (or jump ahead). TL;DR: systems' should never use the wall clock to compute a time interval even if we get rid of leap seconds.
You could simply be working on a single computer and have to deal with time issues. (E.g. Local clock resetting to a prior date and system creating files timestamped earlier than older files.)
I think the leap second is the most minor, not even worth mentioning problem. It is actually to our benefit, reminding us that (perceived) UTC is not a monotonic measure and computer time altogether is not. Computers nowadays adjust their time ~constantly.
That they have to smear their leap second throughout multiple hours, and that computers having the wrong time leads to an outage (for a less than a second difference) troubles me honestly. What you are telling me is that next leap second we should all pray for Meta's NTP servers not going down???
Also: (if you are an fb engineer) What do you do with negative leap seconds? Speed things up...? (https://www.timeanddate.com/time/negative-leap-second.html) The expectation is for a negative leap second coming up -- as the post does mention.
This is an education problem if you expect that time intervals have to be strictly positive (or jump ahead). TL;DR: systems' should never use the wall clock to compute a time interval even if we get rid of leap seconds.