[LEAPSECS] php breaks if UTC has no leap seconds?

[Rob Seaman]

On Dec 10, 2010, at 8:15 AM, Peter Vince wrote:


> I'd be interested if you have some examples of of Y2K bugs that were fixed before they became a problem.  In my very limited experience, I wasn't affected by any, nor aware of them.


When the Y2K inventory was done at the National Observatory, one of our telescopes started to track the sky backwards.

There were a semi-infinite number of fixes to dates in business software and industrial process control:

	http://catless.ncl.ac.uk/php/risks/search.php?query=y2k

For the system I was responsible for (astronomical image processing):

	http://iraf.noao.edu/projects/y2k/y2kplan.html

We set clocks forward, searched 10's of thousands of source files, redesigned key formats and interfaces.

The astronomical image standard had to be changed:

	http://tdc-www.harvard.edu/software/wcstools/fitsy2k.html

(not the normative link, just grabbed from google).

It is a triumph of the astronomical software community that this was accomplished so cleanly, with much credit going to our friend, the late Pete Bunclark.

It's very concerning if Y2K has been recast as a myth among technical professionals, especially among those who were mid-career in the year 2000.  Very significant risks were mitigated through the diligent work of many tens of thousands of software engineers worldwide in dozens of disciplines over years of concerted planning.

Due diligence for a proposed change to UTC would include a similar inventory.

On Dec 10, 2010, at 8:48 AM, Warner Losh wrote:


> You're correct about the question: How many of these interfaces are there, and what is the implication for increasing the error in the present calculation from 1s?  For php, I doubt anybody would care for the cases likely to be in error...  For other things, likely they care more.


Prudent engineering resolves doubts in advance, e.g., "I doubt Hoover Dam will fall down" is an insufficient basis to conclude it is safe.

Rob