Re: Constraints and Functional Dependencies

Marshall wrote:
...
> It is true that the reason behind it all is that we can't check it.
> Here by
> "check" I mean specifically that the technique of doing the
> computation
> for every value in the set being discussed is impossible. (Even in the
> case of, say, 32 bit ints the technique, while not impossible, is
> likely
> prohibitive.)
...

I remember when a USD 500,000 mainframe cpu needed one second to count to 100,000, a good twenty-five years ago. Today a consumer cpu can pretty much count to 32 bit range, either two billion or four billion in that time, even if the loop is written in C and not compiler-optimized.   Surely the prohibition has more to do with presentation or human ability to comprehend the computation.

I'd agree that this comment might be premature if all new consumer cpus several years from now support native integers that are 64 bits long, (which I think is likely because Intel et al don't sell what's best for humanity, just what they've got).

p Received on Sun Feb 25 2007 - 01:55:16 CET