In message <E1Ct1lx-00033E-00_at_mta1.cl.cam.ac.uk>, Markus Kuhn writes:
>You surely must have seen my detailed UTS proposal for how UTC leap
>seconds should be handled trivially and safely by the overwhelming
>majority of computer applications, without any special considerations
>whatsoever by normal application programmers:
>
> http://www.cl.cam.ac.uk/~mgk25/time/leap/utc-torino-slides.pdf
> http://www.cl.cam.ac.uk/~mgk25/uts.txt
Markus, that's all very nice and cute, but it is a grusome hack and
should not be propagated.
It is less gruesome than what we have now in the NTP code, but it
is still a gruesome hack.
There are far to many problems in this that you don't consider:
1) Computers booting inside your 1000 second interval do what ?
2) What about a computer being offline (by design or accidentally)
during the 1000second window ?
3) Many real time systems will not tolerate 1e-3 clock error.
4) How wold a leap-time aware application run on such an operating
system ?
5) You still need to way to distribute leapseconds to embedded and
offline computers.
Your proposal pastes over some of the minor issues with leap seconds,
but it doesn't address the two fundamental problems:
1. You don't know when they will happen with long enough warning.
2. You can't test one when you need to.
--
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.
Received on Mon Jan 24 2005 - 03:22:44 PST