From: Dave Gomboc (dave_at_[hidden])
Date: 2004-01-13 16:49:38
[Dan W. wrote:]
> > Note that if the absolute point happens to be absolute in more than
> > just a relative sense, it may actually enable addition: Adding
> > degrees Celsius or Farenheit makes no sense; but adding Kelvins does,
> > at least in some contexts.
[Phil Richards wrote:]
> I don't see the difference between Celsius, Fahrenheit, and Kelvin if
> you are looking at them in pure absolute (or relative) terms. It makes
> no more sense to add 1K-above-zero-K to 1K-above-zero-K than
> 1C-above-zero-C to 1C-above-zero-C. It makes as much sense to add 1K
> (relative world) to 1K (either absolute or relative) as it does to add
> 1C (relative world) to 1C (either absolute or relative). Both systems
> have a defined zero-point.
I don't see the relevance of a defined zero point here. The important
property that makes any such operations reasonable is that there is an
equality of differences. That is, the difference between 5 kelvins and 6
kelvins is the same as the difference between 25 kelvins and 26 kelvins.
Substitute degrees Farenheit or degrees Celsius for kelvins, and the
statement remains true.
However, there is a difference in another way: switching from degrees
Celsius or degrees Farenheit to using kelvins enables division and
multiplication to be meaningful, e.g. 10 kelvins is actually twice as hot
as 5 kelvins. The systems all have a defined zero point, but the zero
point for kelvins actually represents the absence of heat.
Is there an intent for any of these proposed dimensional and unit
libraries to restrict the types of operations being performed according to
the scale of measurement that the unit belongs to? e.g. 10 kelvins / 5
kelvins gives 2 kelvins, but 10 degrees Celsius / 5 degrees Celsius gives
not only 2 degrees Celsius but also a compiler warning? :-) (An error
would be too severe -- it's pretty common for social scientists to take
means of ordinal data.)
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk