John Cowan wrote on 2000-10-25 20:23 UTC:
> Steve Allen wrote:
>
> > Of course this is pretty much the disease that most systems running
> > NTP currently have.
Yes.
> Not at all. Systems with NTP make smooth adjustments, applicable to
> themselves alone, to a standard that uniformly ticks SI seconds.
That is also the clearly stated idea behind UTS: A standardized
interoperable way of making a smooth adjustment to UTC on systems where
displaying 23:59:60.xxx is not an option and where high-precision
time-interval measurements are anyway neither feasible nor necessary in
practice, because due to interrupt latencies, preemptive scheduling,
variable CPU clock speed, etc. <0.1% time-interval measurement errors
have anyway to be tolerated by applications.
You might want to read the extremely carefully worded proposal again if
this was not evident from a first quick browsing of the text.
http://www.cl.cam.ac.uk/~mgk25/uts.txt
> The proposal would make the very length of the SI second depend on the date!
No, UTS is definitely *NOT* changing the SI second! UTS just defines a
clock display behaviour, not a new physical unit of measurement. UTS can
be converted into UTC and vice versa (using a 1h leap second warning),
so no information is lost. Just like DST is not changing the length of
the SI second to -3599 seconds for one second every fall. Just like UTC
did not change the length of the SI second before 1972. Just like
interrupts delaying a return of gettimeofday() do not change the speed
of light or the meter.
Markus
--
Markus G. Kuhn, Computer Laboratory, University of Cambridge, UK
Email: mkuhn at acm.org, WWW: <http://www.cl.cam.ac.uk/~mgk25/>
Received on Mon Oct 30 2000 - 07:17:51 PST