[LEAPSECS] Terminology question

[Magnus Danielson]

M. Warner Losh wrote:

> In message: <445CB57A-6FCF-4933-A288-BD152135229F at noao.edu>

>             Rob Seaman <seaman at noao.edu> writes:

> : On Mar 10, 2010, at 10:12 PM, M. Warner Losh wrote:

> : 

> : > Maybe we just need to toss the word 'idealized' in there somehow.

> : > 

> : > time_t is an idealized representation of seconds since 1970.

> : 

> : "Idealized representation" is a tautology.

> 

> Actually, it isn't.  If it were an actual representation, it would

> have leap seconds in it.


There are two aspects here, the leap seconds and the actual tracking 
error. time_t tries to achieve two incompatible things at the same time, 
but for all practical matters, implementers chooses to use the mapping 
from the real UTC. time_t isn't looking like either TAI or UTC. It isn't 
piece-wise linear, but it could maybe be said to be nominally piece-wise 
linear to UTC. It's the closest I come right now.

Cheers,
Magnus