[ntp:questions] Re: no more leap seconds?

David L. Mills mills at udel.edu
Thu Aug 11 03:06:41 UTC 2005


You might have heard me honk this agenda before, as I have honked it a 
few times in public and private. We actually have all the pieces in 
place to make everybody happy. Switch all the broadast services, 
computers and NTP to TAI. Done deal. For those applications that salute 
the Sun (bus timetables, astronomers, beekeepers, ...), light up your 
browser and call the International Earth Rotation Service (IERS) in 
Paris and snatch the current UTC (accumulated DUT1) offset from TAI and 
plug it into your sundial, telescope or meeting announcement.

As for NTP, it already does have the capability to download the current 
and historic offsets directly or indirectly from NIST. For compliant 
kernels, ahem, the offset is returned in the ntp_gettime() routine. 
However, that feature depends on all the machines on the way to NIST, as 
well as NIST itself, run NTPv4 and Autokey. Otherwise, just FTP the table.


Thomas A. Horsley wrote:

>>   There is fundamental disconnect between two very different views about 
>>time. One is that time is how long has elapsed from a reference point. The 
>>other is that time is when the sun rises.
> There's not a "disconnect", there really are multiple kinds of time.  That's
> why we have (for instance) both UTC and TAI. Just as people don't tend to do
> long division with roman numerals anymore, they shouldn't be trying to force
> UTC to act like TAI if what they ought to be using in their mission critical
> application is TAI, then that's what they should use.
> What we really need is the addition of a UTC->TAI correction factor to the
> NTP protocol and the addition of TAI based time system services to all the
> popular operating systems. That seems far less intrusive than a complete
> change to what UTC actually means, and could almost certainly be done
> with open source libraries.
> If both time systems are readily available on computers and you still argue
> UTC should change, then what you are really arguing is that all the
> programmers working on all these critical apps are too stupid to make the
> right choice, and if you have such stupid programmers working on this stuff,
> you have bigger problems than leap seconds :-).
> --
>>>==>> The *Best* political site <URL:http://www.vote-smart.org/> >>==+
>       email: Tom.Horsley at worldnet.att.net icbm: Delray Beach, FL      |
> <URL:http://home.att.net/~Tom.Horsley> Free Software and Politics <<==+

More information about the questions mailing list