[LEAPSECS] FaceBook...

[Redman, Russell]

I suspect the real difference is that leap days are extremely in-your-face so everyone is aware of them, whereas leap seconds are so small and seldom discussed until recently that most programmers in the first few decades on the computing industry were not aware of their existence.  

I remember the open contempt expressed for programmers who implemented the full Gregorian algorithm in their calendar functions back in the 1970's, as though their code would still be running 130 years later, hardee har har.  They did this because it was absolutely necessary to implement the every-four-years part of the algorithm for any useful calendar functions, and the 100-year rule was tossed in for hack value.  The 100-year rule would have caused more grief except that 2000 is divisible by 400 and had a leap day, so implementations that did not include the correct rule nevertheless continued to work.  Even completely bogus implementations probably got a look-over in the effort to avoid the Y2K bug.  I wonder, however, whether the Gregorian calendar functions will seem so harmless and well-understood as we approach the year 2100 and the 100-year rule begins to bite the computing industry for the first time.

Unpredictability is a problem that affects leap second implementations, especially for standalone systems, but in principle it is possible to implement the system correctly using the tables of leap seconds from IERS.  It may be possible to arrange that GNSS distribute the required tables periodically, so that even standalone systems can remain up-to-date.  How the resulting BIPM/IERS time-scale relates to civic time in different jurisdictions is an exercise left to the reader.  

The real issue, however, has little to do with leap seconds and much to do with legacy equipment that is difficult or impossible to update, so that companies would rather live with hiccups every year or two rather than face the uncertain consequences of a software update.

Cheers,
Russell O. Redman
________________________________________
From: leapsecs-bounces at leapsecond.com [leapsecs-bounces at leapsecond.com] On Behalf Of Warner Losh [imp at bsdimp.com]
Sent: 09 August 2013 14:06
To: Leap Second Discussion List
Subject: Re: [LEAPSECS] FaceBook...

On Aug 9, 2013, at 3:46 AM, Zefram wrote:


> Warner Losh wrote:

>> beyond any leap day snafus I've seen. Why should leap seconds cause so

>> much more collateral damage?

>

> Maybe that leap seconds are handled in kernel space but leap days in

> user space.


My comments are a higher level: leap days can be implemented without much hassle. Leap seconds have a much higher hassle level. I know why the hassle in the Linux case was higher. My more basic point is that the costs of leap seconds is much higher than leap days, and it tends to be more unpredictable.

Warner

_______________________________________________
LEAPSECS mailing list
LEAPSECS at leapsecond.com
http://six.pairlist.net/mailman/listinfo/leapsecs