
Boost : 
From: Matthew Austern (austern_at_[hidden])
Date: 20010905 11:52:44
"Peter Schmitteckert (boost)" wrote:
>
> Salut,
>
> > > TOPIC: Some "constants" are experimental numbers with finite precision,
> > >
> > > rather than defined values; they may have to be updated when experiments
> > >
> > > get better.
> > >
> > > There seem to be two choices: the values are either compiledin or
> > > read in at runtime.
> breaking compatibility of your results.
>
> > There's also a third choice: assume them to have value of 1,
> > devise your unit system accordingly, and avoid the whole mess.
>
> a small comment:
>
> working with the strange system where c = h/2pi = 1 has the advantage
> that your calculations are independent of measurements of natural constants.
> Suppose you want to build a new atominc clock, which is 1000 times more
> precise than older ones, but to convert the result into seconds you
> have to apply arithmetic with h. You may now face the problem that your
> new device is more accurate than the numerical value of h. therefore
> it might be preferable to work in such a system.
Actually, c is one of the few constants where this is not a problem.
In the SI system the meter is defined in terms of c. By definition,
c is exactly 2.99792458 x 10^8 m/s.
(But yes, in the SI system hbar does have finite precision.)
Matt
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk