>Didi wrote:
>
>> Looks pretty much so. I located a number of UTC clocks
>> on the net and they all agreed.
>> Apparently the NTP time comes modified - so it is not exactly
>> seconds since 1900, but seconds since 1900 excluding the leap
>> seconds... Quite a mess, but since removing the table only makes it
>> simpler,
>> I'll just do it and move on.
>
>This is correct. It may not be elegant but if you stop and think
>about it you'll come to understand that doing it any other way
>would soon become nightmarishly complicated. It would complicate
>simple count->wall time conversions no end in the general case,
>and all machines would either need to know about leap seconds in
>advance (not very likely in the real world) or the protocol would
>need some way of disseminating information not only about upcoming
>leap seconds, but the total number and exact positions of all
>previous leaps seconds.
>
>Added to that you have the need for arbitrary sized tables and you
>can see why the decision was made to fudge it slightly, particularly
>since many (most?) applications don't actually _need_ accurate
>to-the-second timing... Unix time works in a very similar way,
>although the behaviour _during_ leap seconds differs.
>
>In summary, yes it is a fudge, so is the whole concept of leap
>seconds and, indeed, leap years. In the end it doesn't matter how
>you were to define things: it would still be messy.
Eventually we will be able to solve this problem by servoing the rotation and orbit of the earth to our atomic clocks. :)