[LEAPSECS] internet drafts about zoneinfo

[Warner Losh]

On 03/07/2011 15:19, Rob Seaman wrote:

> Warner Losh wrote:

>

>> This mapping is both lossy (because you can't undo it unambiguously) and ambiguous (since the standard insists that leap seconds don't exist).

> "Lossy" isn't quite the right word.  Unfortunately the only antonym offered for "idempotent" is "changed".

>

> The point being that lossy compression may well preserve the essential content of some signal:

>

> 	http://arxiv.org/pdf/1007.1179

>

> while mangling by POSIX does not.

>

> Perhaps "irreversible"?  The mapping is surjective but not injective?  (Depends which mapping we're talking about.)  At any rate it is not bijective, not both one-to-one and onto.


Lossy is the right word here.  It preserves most of the content of the 
signal (defined as an approximation of the value of time) at the expense 
of losing fidelity in some relatively rare events.  It is analogous to 
jpg compression where one preserves the low frequency portion of the 
signal, but smears the high frequency components because the eye has 
trouble telling the difference.

Lossy is also more pejorative than "The function isn't 1-1 onto, so its 
inverse does not exist around the neighborhood of leap seconds." which 
is what I really meant by it.  But that sounds so bland and boring, much 
like "UTC is continuous, but has a non-uniform radix."  Easy to say, but 
it doesn't really convey the pain...

Warner