[LEAPSECS] operational time -- What's in a name?

[Poul-Henning Kamp]

In message <20080328041505.GA20152 at ucolick.org>, Steve Allen writes:


>> In 1961, the task was on a few handfulls of scientific people, most,

>> if not all, of them phd's, and all of them very much at home in the

>> subject domain.

>>

>> Fiddling with time_t today would involves more than a million

>> persons, very few of which can readily tell you what a leap-second is.

>

>I disagree.  All of the technical part falls into the hands of the

>folks who maintain the zoneinfo files and their code equivalents

>(including other, non-POSIX systems for timezone offsets).


No, that is only the part where time_t gets converted to local time.

All the code that assume that time_t is UTC will get burned, and
trust me: that is far more code than you imagine.


>The rest of those million people need merely install a zoneinfo and

>code update as they do every time any jurisdiction decides to change

>the dates of its daylight time transition.


Simply not true.  A good place to start is the FreeBSD ports-tree
which contains about 18000 piece of open source software.

Try to change time_t to a non-arithmetic type and see how far you
get.


>That POSIX time_t henceforth be interpreted as TI, and that the leap

>seconds go into the zoneinfo files such that most time zones,

>including UTC, start being a number of hours minus leap seconds off

>from TI.


This is exactly the flagday that will make the upgrades to a few
hundered telescopes look like peanuts.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.