[time-nuts] 15 Seconds error...??
magnus at rubidium.dyndns.org
Wed Jan 18 16:36:08 EST 2012
On 01/17/2012 06:24 AM, Jim Lux wrote:
> On 1/16/12 6:58 PM, Magnus Danielson wrote:
>> On 01/17/2012 03:18 AM, lists at lazygranch.com wrote:
>>> I think the time jumps would be a problem for the GPS "customers".
>>> Suppose I am computing speed by differential readings. The jump would
>>> be an issue, especially if time went backwards.
>>> File stamps. Stock timing. Similar problems.
>> If you do not desire jumps, do not use UTC time but GPS time out of the
> or TAI...
> Jumping the internal time-scale would upset things, so a
>> none-jumping time-scale is used. If an application requires a
>> non-jumping time-scale, then use GPS time-scale for that application.
> I just had an discussion last week with people who wrote a spec
> requiring GMT (copying an older spec that antedated UTC, I suspect).
> Subtle details of GMT vs UTC are lost on them. When the leap second
> notice came out I asked what they were going to do (since the logs need
> to be monotonic and continuous).
> For now, they'll use the hack of shutting down before UTC midnight on 30
> June and starting up later.
> It turns out that most of the system actually uses TAI/GPS time and
> hence it isn't a problem, but it's technically non-compliant with the
> requirement for GMT. So they're stuck:
> 1) admit the system does not meet the requirement and seek a waiver of
> the requirement
> 2) Change the requirement
> Both are painful.
> Life would be easier if they just said.. use TAI time.
Use of TAI is fine, as long as you won't tell anyone near the timelords,
as obvious TAI is supposed to be "hands off".
I think they should change the requirement.
More information about the time-nuts