Re: Impossible Database Design?

> On Monday, 7 October 2013 14:59:16 UTC+11, David BL wrote:
> > On Monday, September 30, 2013 11:37:06 AM UTC+8, Derek Asirvadem wrote:
>
> I was prompted to read that "Impossible Database Design?" thread from 7 years ago, and found something Bob Badour said which was unchallenged at the time. Better late than never I guess...
>
> > With all due respect, inifinite is about capacity and the impossibility
> > of representing each of an infinite set of values in a finite space.
>
> For this statement to be correct it must be assumed the (non)existential quantification over finite spaces is outside the universal quantification over the values of the infinite set.
>
> I believe in the context of the somewhat superficial discussion (about finite versus infinite), the statement was intended to suggest that infinite data types are not useful to computer science. If so I'd call that a gross misrepresentation!

<snip>

Yes, I agree.

More important, while Bob's comment may be worth an academic argument, it is not relevant to this thread, Nikolai's requirement. If you read and understand my solution: https://groups.google.com/d/msg/comp.databases.theory/Rhv-9vZcj48/pkAwy3ScxmoJ one has to use a Projection in SQL (not the relational calculus projection). So the calendar may well be infinite, from the users perspective. But on a finite computer, the source used for the Projection is finite, but the limit of that is nowhere near the limit of a finite value, or the end of a range that is close to infinity, etc. Infinity itself is not required to be stored, neither is the far reaches of the end of the rage that is close to it. One company may load ten years worth of dates, another company may load 100 years worth of dates. Both are tiny vis-a-vis the finite limit of whatever key you are using.

> TTM has a prescription that all types are finite. I find that completely unreasonable (baffling really). I discussed this on TTM mailing list and people were divided on the issue.

TTM is not worth either my time or yours. Nothing is ever resolved or produced. Only endless, unresolved arguments about problems that the real world does not have.

I agree that all types are finite. Codd states that, refer his declarations re domains. That is not arguable.

What they did argue, during my three years there, was the representation of it. Much like Bob's assumption that somehow, the representation is limited to the finite computer, but TTM-ers were sillier. Maybe they have not heard of the manner in which floating point numbers are represented in a computer. So yes, the range is finite *in a computer*, but with a logarithmic representation, plus plus, representing (a) very large ranges and (b) infinity, is a non-issue.

Cheers
Derek Received on Tue Feb 25 2014 - 15:19:16 CET