On Jun 15, 2007, at 1:57 PM, Poul-Henning Kamp wrote:
However, for the approx 200 computers I have in my personal sphere
of influence, I can in less than 1 second split them into these two
categories:
1. Does this computer care if there ever is another leap-second ?
199
2. Does this computer care if civil time is aligned with earth
orientation ?
1
Yes, yes - operating systems and embedded hardware are one place to
look for DUT1 (Y2K-like) dependencies. So too are applications,
library code, interface definitions, GUIs, network dependencies.
This particular discussion was started with concerns about DoD
assets. Might modules, interfaces, libraries, implicit data
structure interdependencies, human factors, operating procedures and
so forth comprising the system of systems that is a ballistic missile
submarine, or cruise missile, or Pacific fleet operations require
that heretofore negligible DUT1 corrections be interposed?
Both logic and prudence suggest yes. Having invested in such a USN
or DoD or NASA wide inventory, one will either discover that
(unlikely as I believe it to be) no code or operational changes are
needed - or that changes are indeed needed. If changes are needed,
the updated systems will require lengthy and expensive formal
testing, integration, deployment. But perhaps less obviously, even
if no changes are deemed necessary an open-ended regime of testing
and retesting will also be needed to uncover any Y2K-like issues
triggered as DUT1 grows secularly and crosses code-dependent
boundaries. Effects will be subtle to discover, but perhaps - umm -
"dramatic" in their implications.
And can you really be so confident that 99.5% of your systems are
100% free of any possible DUT1 related issues? Microsoft can't even
get daylight saving time correct. We've already established that
different countries rely on different legal notions of time. Do your
computers talk to any that reside in other countries?
The bumper sticker bottom line:
Absence of evidence is not evidence of absence.
Rob Seaman
NOAO
_______________________________________________
LEAPSECS mailing list
[email protected]
http://six.pairlist.net/mailman/listinfo/leapsecs
