[time-nuts] Leap second? Yay or nay?

[Majdi S. Abbas]

On Mon, Jul 02, 2012 at 10:24:33AM +1000, Jim Palfreyman wrote:
> As an astronomer I've been a supporter of the current leap second situation
> and have not really liked the idea of changing.

	As someone who still owns a sextant and almanac, I concur.  :)

> However, after yesterday I'm thinking of changing my mind.

	Understand that yesterday's situation was specific to one
operating system, which accepted a patch to its kernel a few years ago
that was never really tested.

	What happened was more of a condemnation of bad software and no
testing.  FWIW, I spent most of today cleaning up after it, and I still
feel this way.

	This could have been extensively tested.  It wasn't.  It's not
hard to simulate a leap, either via ntp, or directly via mechanisms like
adjtime().

	Finally, if people really don't want to deal with leaps, they
are welcome to use TAI.  Why have two atomic timescales that don't
observe the leap, offset a fixed number of seconds from each other?

> This could also have been serious. Were planes tested in-flight for this? I
> bet they weren't.

	Speaking for myself, I was on Southwest 2449 during the leap,
and wasn't for a moment concerned about it.  I trust the flight crew to
do their jobs.  

> Software writers the world over are notorious for not fully testing their
> code, so the leap second situation in our increasingly time-dependent world
> has the potential to one day take a life.

	I can think of a lot of other ways bad code could kill someone (
utility control systems, heart monitors, automotive control, etc.)

	I don't want to see anyone hurt, but pandering to bad software  
just gets you more bad software.  At some point quality has to matter.

	--msa