Re: Temporal database - no end date

On Jan 18, 9:14 pm, "-CELKO-" <jcelko..._at_earthlink.net> wrote:
> >> All of your continuum-based arguments are irrelevant to digital
> > computers; they can't handle a continuum anyway.
>
> Things like pi and e are handled in many programming languages with
> tokens, much like the NaN.

Some computer algebra systems work this way, but programming languages? I can't think of a single one. (And I can probably name more programming languages than most people you know.) In programming languages we pretty much universally use the nearest double precision floating point number instead of an abstract token. And that usually gives us a lot more precision than we're going to need.

> **Symbolic** manipulations like you do on
> paper are just as good as computational ones!!

IEEE floating point numbers *are* symbols. You probably meant algebraic manipulations. And I wouldn't say they are "as good"; they are better, if you are talking about precision. However for the vast majority of use cases, the extra precision isn't worth the trouble.

> Your mindset is
> limiting the computer! Think about it.

I have thought about it plenty, and my mindset is not the problem. The difficulties in handling a continuum are fundamental and intrinsic, and not a failure of imagination.

> There are two ways to define a set: (1) enumeration
> (2) characteristic function.

Only two?

nat = zero | succ(nat)

> This is important. Enumeration must list ALL of the values --
> impossible in an infinite set. Characteristic functions must always be
> able to return TRUE or FALSE .. well, there are set that cannot have
> such functions, but ingore the graduate level math for now that puts in
> the 3VL that Chris Date hates so much
>
> The half-open interval is a (2) method and the Chronons are a (1)
> methods. Let's say that we have one DB with a precision of days like
> Chris Date and another DB tha tworeks on weeks start on Monday. What
> do we do about Tuesday? In model (1), we freak out and say there is no
> such element in this domain. In model (2) we see that it falls after
> Monday and before Wednesday because we have a test.

Nothing in the last two paragraphs has any relevance to modeling continua that I can detect. Again, we model the reals with floating point, and it works just fine, even though the reals have the power of the continuum and floating point doesn't. If we want to bring it back to time, then we have to consider that time in the computer has a machine-defined resolution, and that makes time in the computer discrete, regardless of anything Einstein said about the real world. Java models time with a 64 bit int, and that gives you millisecond resolution over half a billion year span. That's good enough for my apps.

Marshall Received on Fri Jan 19 2007 - 06:37:24 CET