[LEAPSECS] Leap smear

[Rob Seaman]

On Sep 26, 2011, at 4:03 PM, Clive D.W. Feather wrote:


> No, because NPL has agreed to keep that clock running at two rotations per

> 794243384928000 periods of the radiation corresponding to the transition

> between the two hyperfine levels of the ground state of the cesium 133 atom.


Specification, not requirement.


> I actually have no problem with your call for engineering requirements.

> Though I'd like to know where they were when the leap second was invented.


Oakmont elementary school.


>> These assertions are both de jure, both specifications.

> 

> Not for what "de jure" means to me. The latter definition is one of

> convenience to me; it is *not* one of law in any shape or form.


Whatever human law means to you, it is a mechanism for achieving goals in a world of natural physical requirements.


>> The underlying requirement is not something like "thou shalt honor mean solar time at Greenwich".

> 

> Quote correct. Glad to see you finally admit this isn't a requirement.


Yet another poor gangly straw man knocked down.  It is, however, the current specification.  And the underlying requirement is not only deserving of more respect than expressed, but it is indeed "required".  It won't go away through word games.


>> Another way to state the underlying requirement is that calendars count integral days.  Points awarded for anybody who can make this work for some definition of day that does not remain stationary with respect to mean solar time.

> 

>    '"day" is defined as 794243384928000 periods of the radiation

>     corresponding to the transition between the two hyperfine levels of

>     the ground state of the cesium 133 atom.'

> 

> What was hard about that?


Nothing hard.  Simply not a requirement.  It is a blatant specification that is currently inaccurate.


>> An engineering requirement is not a specification.

> 

> Agreed.

> 

> However, nor is it an axiom. There's nothing magically correct about the

> synodic definition of "day", any more than there's anything magically

> correct about the TAI or UTC definitions of it.


Magical, perhaps not.  Inherent in the context between humanity and the reality we inhabit.  Yes.


> Which we choose to use for civil time in any particular place is purely

> a question for the politicians.


Nonsense.  (And pretty scary nonsense considering some of the politicians.)

However, consider this assertion.  Unless you are really intending to state that politicians can literally decide to spin clocks backwards, change the rate by orders of magnitude, or turn clocks over, under, sideways, down - then there are engineering requirements hidden in statements like this.  Politicians - even politicians - have constraints on their actions, perhaps especially on wonky technical issues like timekeeping.

The ITU proposal is only even notionally possible for consideration because some deem it close enough for government work.  The thing it is deemed to be close enough *to* is precisely mean solar time.  This is what it means to be an engineering requirement.  What is "LOD" in all those plots?

However, as with horseshoes, the question of close enough has implications.  Engineering best practices would suggest that consensus on how these implications will be addressed be reached before action is taken.  Rather than bringing order, the ITU proposal unleashes anarchy.  Without a coherent plan in place beforehand, random non-governmental forces will come into play afterwards.

Rob