[boost] Re: Re: Re: Fixed-decimal partial review

From: Misha Bergal (mbergal_at_[hidden])
Date: 2003-07-21 23:27:07

"Beman Dawes" <bdawes_at_[hidden]> wrote in message
news:4.3.2.7.2.20030720203206.01eafad8_at_mailhost.esva.net...
> At 05:13 PM 7/20/2003, Misha Bergal wrote:
>
>  >Something seems to be wrong to me tough. The programming languages
>  >(environments) introduced lately do not implement fixed-point only
> decimal
>  >numbers.
>
> Fixed-point is used in applications driven by external requirements, not
by
> what is or isn't available in programming languages.

I don't think people who did Java or C# are ignorant of requirements of
financial applications. They must have had the reasons for doing floating
point instead of fixed point only.

> That's possible, but my guess is that it is going to be a long time before
> floating-point becomes acceptable in the broad world of business,
industry,
> and government.

Do you mean that it is unacceptable because of the difference between 1) and
2):

1) Fixed point (2 digits after decimal point) :
1.00 + 1.00 * 0.06 * 300.00  = 1.00

2) One of the implementations using floating point:
1.00 + 1.00 * 0.06  * 300.00  = 19.000000

Well, at least 2) explicitly shows that the necessary rounding did not
occur, in case of 1) it is not so explicit, so I would not be sure that 1.00
is the intended result.

> Try it on your tax return and see what kind of a response
> you get:-?

When calculating my tax return I use the same arithmetic my calculator uses.
I believe it uses floating point.

> Uh... In the applications I have in mind it is floating-point arithmetic
> that has a bad reputation for producing technically correct answers that
> are dead-wrong in terms of the application's needs.

Do you have any references regarding bad reputation of floating point
arithmetic that I can look up?

Misha Bergal
MetaCommunications Engineering