On 2026-10-16 10:56, Guy Harris via tz wrote:
Could you, or one of your human minders, please add "don't write fancy HTML...
Yes, plain text is preferred. That email was "Content-Type:
multipart/alternative" and (as is my practice) I looked only at the
text/plain alternative, for the usual security reasons.
It's primarily of interest to the aforementioned director, so presumably Paul
will forward your message to him.
Yes, I forwarded it, along with the following additional comments.
Thanks for the report.
----
AI agents at CyberNative AI LLC surveyed nine downstream users of
leap-seconds.list, and found that all of them accepted an expiry date of
2036-01-01, but one partially failed with 2100-01-01. The partial
failure was due to the NTP wraparound, which occurs 2**32 seconds (not
counting leap seconds) after 1900-01-01 00:00:00 UT, i.e., the last
valid second before the wraparound is 2036-02-07 06:28:15 UTC, and the
wraparound occurs one second after that.
So, it might be useful, at least for now, to pick an expiration time of
2036-02-07 06:28:15 UTC or earlier in leap-seconds.list, along with a
comment that this expiry is temporary and is only to provide some time
of transition away from classic ntp 4.2.8p18. After some delay after
classic ntp fixes the problem (or if 2035 rolls around, whichever comes
first), you can then substitute 9999-12-31.
Although I filed a bug report against classic ntp here:
https://bugs.ntp.org/show_bug.cgi?id=4042
they are in legacy maintenance mode and haven't issued a release in more
than two years. Luckily classic ntp is not used much nowadays.