[LEAPSECS] Royal Society

[Markus Kuhn]

Steve Allen wrote on 2011-11-08 17:27 UTC:

> http://www.nature.com/news/2011/111108/full/479158a.html

> 

> "they failed to reach a consensus"


Just for the record: the Nature article may give the impression that I
said in the phone interview something along the lines of "Linux
operating systems, for example, have experienced problems because they
add the whole second in one abrupt jump at midnight, which confuses the
software." In fact, I only said that this abrupt implementation of the
leap second in the current Linux kernel *could* in theory cause problems
because it makes computer time non-monotonic. I never said I know
evidence of noteworthy real-world problems caused by Linux-style leap
seconds. After all, misbehaving clocks are a common feature of the real
world, and applications need to deal with them robustly in any case,
with or without leap seconds. I believe that leap seconds cause only a
tiny fraction of abnormal clock behaviour seen in real-world systems,
and if they stand out then only because they are widely advertised in
advance.

In a discussion where the collection of evidence for leap-second caused
real-world problems resembles the game "Chinese wispers", I'd rather not
be seen contributing to the latter.

I was also quoted as supporting the Google solution for implementing
leap seconds. That's an oversimplification: for a more detailed review
of Google's approach versus the UTC-SLS proposal, see

  http://www.cl.cam.ac.uk/~mgk25/time/utc-sls/#google

Markus

-- 
Markus Kuhn, Computer Laboratory, University of Cambridge
http://www.cl.cam.ac.uk/~mgk25/ || CB3 0FD, Great Britain