[LEAPSECS] The relation between calendars and leap seconds.

[Rob Seaman]

Tony Finch wrote:


> On Wed, 12 Nov 2008, Rob Seaman wrote:

>>

>> However, surely the point of coupling TI with the zoneinfo notion -  

>> just for

>> the sake of argument - is to simply start distributing TI instead  

>> of UTC.

>> Then UTC - a flavor of Universal Time, an alias for Greenwich Mean  

>> Time (which

>> as you say is far from "eradicated") - remains available for those  

>> who need

>> it.

>>

>> The adoption of TI would be massive and immediate.

>

> If you simply do that then civil time will de facto diverge from UT,  

> since

> nobody will be applying the TI-UTC offset to the time broadcasts.


No, the realization of civil time will diverge from UT until bad  
engineering is fixed in the fullness of time.

Everybody agrees we can cheat mean solar time for many purposes for  
some number of years.  For those purposes that need correct UT  
behavior immediately (meaning correct UTC behavior as your use of both  
terms neatly illustrates), more aggressive pressure will be applied to  
find appropriate mechanisms - whether zoneinfo or not - to recover  
UTC.  Other systems will follow as the distinction between TI and UTC  
grows.

However, the initial transition to TI simply requires cutting a ribbon  
with big scissors and declaring it to now be so.  I'm a bit perplexed  
at how astronomers are at fault if we don't consider such options and  
are still at fault if we do.


> Remember that (as previously discussed on this list) you can't simply

> redefine time_t to be TI and fix up the difference using zoneinfo.  

> This is

> because plenty of software doesn't use zoneinfo for time  

> computations, and

> instead relies on the POSIX specification that time_t is roughly UT.


Depends on the meaning of "fix up".  I would take it to mean actually  
using technology in an appropriate fashion.

Are people really convinced by the argument that badly implemented  
systems should determine policy?


> You have for many years been able to configure the tz code (and  

> therefore

> many unix systems) to run with time_t following TAI and with the TAI- 

> UTC

> fixup in the zoneinfo code, just like the proposed TI setup. No-one  

> runs

> it this way because it isn't conforming and therefore isn't compatible

> with a lot of code.


If the solution we're discussing came to pass through the fullness of  
time and the application of systems engineering best practices, then  
such usage would become conforming and the balance of code  
compatibility would shift.

As the consensus in Torino recognized, TI is one of the more gentle  
transitions to pursue.

For the ITU to win, why must UTC lose?

Rob