Re: Temporal database - no end date

From: Aloha Kakuikanu <aloha.kakuikanu_at_yahoo.com>
Date: 23 Jan 2007 18:35:16 -0800
Message-ID: <1169606115.973286.102490_at_a34g2000cwb.googlegroups.com>


On Jan 23, 6:05 pm, "David" <davi..._at_iinet.net.au> wrote:
> On Jan 24, 3:29 am, "NENASHI, Tegiri" <tnmai..._at_gmail.com> wrote:
>
> > One can not model the continuum with the floating numbers because they
> > are discrete and finite like the computer integers.
> Evidently you require models to be isomorphic to what they model. I
> presume you would say an C++ "int" doesn't model an integer, or a
> computer doesn't model a universal Turing machine. I have a far less
> restrictive notion of the word "model".
>
> How would you describe the relationship between floating point numbers
> and the reals? Pretends? Imitates? Masquerades? We need some word.

Bastardize?

I agree with Tegiri that int is much more satisfactory model of integers, compared to float model of the real numbers. It is a common knowledge that the only field where floats are useful is computer graphics.

Peter Potts has written a very nice thesis how floating point arithmetics should be done via continued fractions approximations: http://www.purplefinder.com/~potts/exact/phd.pdf Received on Wed Jan 24 2007 - 03:35:16 CET

Original text of this message