[LEAPSECS] Crunching Bulletin B numbers (POSIX time)

Warner Losh imp at bsdimp.com
Wed Feb 23 13:13:36 EST 2011

On 02/23/2011 03:59, Ian Batten wrote:

>> Nope. tried that when getting the spec approved. Approximate times weren't allowed. UTC times were required. There was no way to indicate approximate time in the user interfaces present (how do you blink a 5071A anyway :). The other systems that interfaced to ours had a fixed format, and required UTC and not approximate UTC for a while and then a possible jump in time to actual UTC.

> Well, if the use-case is navigation (Loran, military) then UTC sans leap seconds isn't much use to you anyway, so the "solution" of dropping them would take your requirement with it, and you'd seen something closer to UT1. And of course, requirements are simply line items on an invoice, and if deriving immediate UTC costs 10X rather than X, the customer has to make a decision on whether they're prepared to pay for it. Just saying "my customer demands X and therefore the rest of you need to enable X" isn't realistic.

you've never dealt with the military have you? Our customer demands
what they demand. We try our best to talk them out of crazy
requirements, but with the military that's not always possible. They
must comply with the dictated requirements in some cases from higher
up. UTC shall be used for all timestamps. All errors in time keeping
must be logged with timestamps. Boom, they are cross coupled...

Really, I tried really hard to decouple the UTC requirement from the
timekeeping requirement, but there were too many cross couplings to do
that in a cost-efficient manner. So we did the best we could with the
data available, but since the customer tended to focus on worse case
scenario, even extremely low-probability ones, we had to cope with that.

There was also no source of UT1 available, so that's a non-starter. The
only time we punted on the UTC requirement was to log that UTC was not
available, and we timed out waiting for it. And that put the system
into a holding pattern until UTC could be obtained.

And yes, there are multiple, redundant sources of time around. And we
did use those to make clever guesses. But in the end, that didn't help
the worst case scenario at all.


More information about the LEAPSECS mailing list