[LEAPSECS] June 30 not leap

Warner Losh imp at bsdimp.com
Tue Jul 2 11:12:02 EDT 2013

On Jul 2, 2013, at 3:49 AM, Zefram wrote:

> Warner Losh wrote:

>> It is gratifying to see how flawlessly this standard is implemented,

>> despite over 40 years of it being the standard. I guess the far older

>> expectation that every minute has 60 seconds is kinda deeply engrained...


> Wikipedia drily notes that "Between 1000 (when al-Biruni used seconds) and

> 1960 the second was defined as 1/86,400 of a mean solar day". That 1960

> endpoint to which it refers is the transition to the ephemeris second.

> Rubber-seconds UTC, which has occasional minutes with !=60 seconds,

> comes along in 1961. So we've got around 960 years of the second only

> being a subdivision of the minute, versus 53 years or so of seconds

> being more complicated than that.

And only about 10-15 years where we had computerized time keeping that was sufficiently accurate and connected to notice leap seconds. In college (1985-1989), leap seconds may have received a quick mention "well, pedantically time does this, but nobody bothers with that stuff in computers" to seeing mentions on risks in the early 1990s to seeing GPS receivers come down in price and the internet explode in late 90's early 2000's, to having the 7 year dry spell between 1998 and 2005 where there was talk about just eliminating them altogether. So people really have only had three leap second disruptions (and a couple of non-leap-second disruptions when we didn't have one after fixing the bugs the last real one showed up). There's a lot of time-keeping code out there that's been written that just doesn't care, and sadly that code limits one's ability to get leap seconds right in real time... The past 8 years where it has started to really matter hasn't been enough time for it to sink in. Otherwise Google would do rubber seconds around leap second events, for example...


More information about the LEAPSECS mailing list