On Thu, 9 Apr 2020 22:33:09 -0400, Bob Bridges wrote: >I just read an article about time-keeping that made a claim that puzzles me: > >"But most people — including commercial programmers, who write the critical >software that controls public and private infrastructure — don’t know about >the leap second, Matsakis said, and that means their code doesn’t account for >it. So when a new leap second rolls around, things break. Reddit, LinkedIn, >and Yelp all suffered issues related to the last leap second in 2012. And, >more seriously, computer booking systems used by Qantas Airlines all >struggled, delaying flights by hours. > >"In some cases, it is impossible to update systems before the next leap second >arrives. Matsakis spoke of a Switzerland power company whose backup systems >only turn on when needed—otherwise, they sit disconnected from the network. >When they were activated in a test after the last leap second, they crashed." > >What's this about? What would crash because we didn't account for a leap >second? Why wouldn't a plane scheduled to leave at 13:44:20 on a certain day >simply leave at 13:44:19 instead? > Sometimes validity checks reject a transaction that appears to have completed before it started. If the transaction was critical, the consequences may propagate. Error injection and fuzz testing may not have discovered the weakness.
See: "leap smear" in https://en.wikipedia.org/wiki/Leap_second#Workarounds_for_leap_second_problems See (my favorite): https://www.nist.gov/pml/time-and-frequency-division/time-services/ut1-ntp-time-dissemination Just do it! https://datacenter.iers.org/data/latestVersion/16_BULLETIN_C16.txt >The article is here: >https://www.theatlantic.com/technology/archive/2015/01/the-over-perfection-of-humans-global-clock/384355/ -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
