Re: floating point

[Jon Saperia <saperia@jdscons.com>]

Andrew, thanks for good ideas. Sometimes a fixed approach such as you
describe can work. There are some places as in the case of latencies
through a network over time where floating point values are better -
there can be orders of magnitide change overtime (the dynamic range
problem). This cascades into the derrivative objects based on these
values such as means and standard deviations. I may use some of your
ideas, but in the end my view is that we are past due for floating point
numbers.

/jon


> Jon,
> 
> Before deciding to go there, I would first check your initial premise that
> you "have to" define some floating point objects. Maybe there's a simpler
> kludge that is sufficient e.g. a) define the units in 10ths, 100ths or
> 1000ths; b) define as 1/x where x is an integer; c) define as an exponent
> 2^^x or 10^^x; d) any combination of the above. Such kludges have usually
> been sufficient.
> 
> Of course I'm sure you've done that and have valid reasons for needing
> greater dynamic range or something, but using true floating point is quite a
> leap (and arbitrary fixed point arithmetic ought to be simpler than floating
> point but often turns out not to be due to lack of suitable libraries or
> other reasons). And if this debate has already been had on the SMIng list
> then my apologies for wasting time/space.
> 
> Andrew
> 
> 
> -----Original Message-----
> From: owner-mibs@ops.ietf.org [mailto:owner-mibs@ops.ietf.org]On Behalf
> Of Jon Saperia
> Sent: Tuesday, October 30, 2001 5:32 AM
> To: mibs@ops.ietf.org
> Subject: floating point
> 
> 
> I have to define some objects as floating point and can not wait for the
> sming work to be completed. Is there a convention that people have used
> such as two objects with one indicating the number of decimal places?
> 
> 
> Thanks,
> /jon
> 
> --
> Jon Saperia		     saperia@jdscons.com
> 			     Phone: 617-744-1079
> 			     Fax:   617-249-0874
> 			     http://www.jdscons.com/
> 
> 

Thanks,
/jon
--

Jon Saperia		     saperia@jdscons.com
			     Phone: 617-744-1079
			     Fax:   617-249-0874
			     http://www.jdscons.com/