Re: Character string relation and functional dependencies

From: V.J. Kumar <vjkmail_at_gmail.com>
Date: Sun, 9 Dec 2007 05:05:25 +0100 (CET)
Message-ID: <Xns9A00EAF626A84vdghher@194.177.96.26>

Tegiri Nenashi <TegiriNenashi_at_gmail.com> wrote in news:cd2ad8bd-78ca-433d-bc0f-3e7ef0c0fe2a_at_e6g2000prf.googlegroups.com:

> On Dec 6, 2:38 pm, Jonathan Leffler <jleff..._at_earthlink.net> wrote:

>> Tegiri Nenashi wrote:
>> > On Dec 6, 9:40 am, rp..._at_pcwin518.campus.tue.nl (rpost) wrote:
>> >> Another difference is that database tables are finite and
>> >> variable, 
>>
>> > Oh, relations in database world are certainly not restricted by
>> > finite cardinality.
>>
>> I thought that computers are finite, so the relations containable in
>> them are too - even if damn large.  There's a big difference between
>> very large and infinite.

>
> This doesn't really matter. You can still reason about infinite
> relations

You can do that with your brain...

with finite resources available on you computer platform.

but not with that. The computer is an intrinsically finite gadget. Therefore, you'd better use the finite model apparatus to reason about things like the impossibility of expessing transitive closure in the relational algebra. A lot of stuff like the compactness theorem does not work with finite models which makes infinite model proofs inapplicable in the finite case.

>

>> One ultimate limitation is the uniqueness requirement.  Suppose you
>> have a table with two integer columns.  Since the range of the
>> integer types are finite (even if your DBMS handles multi-precision
>> integers), then the maximum number of distinct rows in the relation
>> is also finite. 

>
> All computer algebra systems work with numbers which are not
> restricted by a whim of hardware architects. 16/32/64 bit integer
> numbers (let alone floats)? give me a break!

Jonathan is of course right, the set of 'floats' is clearly finite that somewhat clumsily approximates real numbers !

>
Received on Sat Dec 08 2007 - 22:05:25 CST