On 2015-10-28 19:12, Timothy Sipples wrote:
> 
> A leap second "spin" is one *option* customers have. It's not a
> requirement. The other option available is ... gradual steerage to adjust
> for the leap second! It takes approximately 7 hours for the IBM z System
> and z/OS to smooth out one leap second. Refer to IBM Tech Doc # WP102081
> for more information:
> 
> https://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/WP102081
> 
Thanks.  I hadn't been aware of option 2 (that Tech Doc is a mere 4
months old.)

> So, unlike Google and Amazon, *you* choose. If you want precise timing at
> every moment, choose that option (not available with Google and Amazon). If
> you want gradual steerage, choose that. Unlike Google and Amazon, IBM does
> not force one choice on its customers here. And you can change your mind
> before the next leap second if you wish.
> 
Choice isn't always a good thing.  The saying, "The wonderful thing about
standards is you have so many to choose among!" is intended sarcastically.
And:  
https://www.ted.com/talks/barry_schwartz_on_the_paradox_of_choice?language=en

And the Tech Doc doesn't address the Google/Amazon practice of *centering*
the steering on the leap second, so they're never over a half second away
from UTC.

> Come to think of it, there's probably a third option: do whatever you want
> on your external time reference, ...
> 
OK.  RYO.  You could do it that way.  Or perhaps use Google as the ETR source:
one standard fewer.  (Amazon steers for 24 hours; Google for 20.)  UTC is
an unfortunate choice for IT; UT1 would have been less disruptive, thus
wiser.  Inertia precludes a repair.

Looking for a round number, a second in 7 hours is very close to 40 ppm.
But that must also compensate for any intrinsic error in the CPU's
crystal (? I guess.  OCO?) clock.  Suppose that internal reference is
10 ppm fast (still better than Timex. Is there a spec for this?)  In that
case, the effective steering rate is 30 ppm, and it will take about 9
hours rather than 7.

Thanks again,
gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to