[LEAPSECS] Toasting Unix timestamp 1234567890

[Magnus Danielson]

Zefram skrev:

> Magnus Danielson wrote:

>> Indeed. This is useful for those believing in SI/TAI seconds for as the 

>> second distance of time_t, their reference epoch is inconveniently 

>> offset in this fashion, making them have an offset of 25,999918 s from 

>> POSIX (right now - propper reference scale should be TAI).

> 

> Yes.

> 

>> Those believing in rubber seconds up to 1972 and then SI/TAI seconds 

>> from that will get the more convenient 10 s offset from 1 Jan 1972, 

>> making them having an offset of 24 s from POSIX.

> 

> No, if you're doing this then you must include the irregular leap.


Sorry, I think you over-interpret a poorly articulated formulation of 
mine. If you think according to "Honour the UTC definition from 1970 to 
1972 and then the new leap-second based UTC definition from 1972 up to 
current time" then I think you should come to the same conclusion as I 
do. The honouring of the UTC definitions includes the leaps in offset 
both during the pre-1972 definition as well as at the time of change to 
the new standard.


> The current offset is about 24.107757997 s, and does not have a terminating

> decimal representation.  (This is the counting-UTC-seconds way.)


I do not understand what that number comes from. Does not match what I 
meant at least, so you need to describe what it means.


> I believe some people do do the 24 s offset that you suggest, however.

> I think it's through ignorance of the rubber-seconds era and its leaps.


So true.


> These people are effectively using an epoch of 1972-01-01T00:00:00 UTC

> = 63072000.  There's no clean description of what they're counting since

> 1970.  If doing this then it seems more consistent for pre-1972 timestamps

> to count back in TAI seconds, but in practice I think a vague-UT pre-1972

> timescale gets grafted onto the TAI-synched post-1972 timescale.  Bit of

> a mess.


It is a bit of a mess. Honouring the pre-1972 UTC definition for the 
pre-1972 era makes sense as it allows for a practical solution at least, 
as only integer offsets is involved. The POSIX re-defintion patches the 
apparent drift between time_t and UTC datums caused by leap seconds, but 
that is a post-1972 era issue and does not fully explains the pre-1972 
era behaviour, but could be interpreted to honour the pre-1972 UTC 
definition, but then the reference epoch is not occuring at the same 
time as the original UNIX GMT defined epoch.


>> 1234567890 seconds of time_t makes no sense at the time that time_t 

>> reads 1234567890 since it is not the number of seconds from the 

>> reference epoch, it is a form of "mock seconds" to make the scales fit. 

> 

> Yes.  This is sometimes succinctly described as counting "non-leap

> seconds since the epoch".


Well... that only explains the behaviour in the post-1972 era, where as 
it does not really help for the pre-1972 era where leap seconds was not 
used, so even that wording does not sufficiently covers what is 
happening unfortunately.

At 1 Jan 1972 we jumped from 9,892242 to 10 s offset in one fractional 
step. Before that we had a 3E-8 s/s phase ramp from 8,000082 of 1 Jan 
1970. Neither qualify as leap seconds.

Cheers,
Magnus