[LEAPSECS] [time-nuts] Leap Quirks

[M. Warner Losh]

In message: <alpine.LSU.2.00.0901051603040.10002 at hermes-1.csi.cam.ac.uk>
            Tony Finch <dot at dotat.at> writes:

: On Mon, 5 Jan 2009, Poul-Henning Kamp wrote:

: >

: > The proper thing for the future is either a "int128_t" 64.64

: > fixedpoint time representation or a double ditto.

: 

: Do you mean double as in the C type? Which is surely too small - you want

: quad-precision FP or perhaps "double double" (paired doubles to get 112

: bits of mantissa).

: 

: Or do you mean 256 bits of precision, in which case, boggle!


I think he means int128_t, for those systems that implement it.  A 128
bit integer that's interpreted by the time routines as a fixed point
number.  64 bits of second offset, and 64-bits of within the second
offset.

Warner