Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Usenet -> comp.databases.theory -> Re: atomic

Re: atomic

From: Bob Badour <>
Date: Wed, 31 Oct 2007 20:07:18 -0300
Message-ID: <47290aa8$0$14836$>

paul c wrote:
> Bob Badour wrote:

>> paul c wrote:
>>> David BL wrote:
>>>> On Oct 31, 4:31 pm, "Roy Hann" <specia..._at_processed.almost.meat>
>>> ...
>>>>> 1NF does not *require* that values be atomic.  It asserts that 
>>>>> values will
>>>>> be *treated as* atomic.  Big difference.  Essential difference.
>>>>> Roy
>>>> Can that be formalised?  I agree with Bob that in general we have a
>>>> set of operators and they can allow us to see internal structure.
>>>> What does it mean for a value to be *treated* as atomic?
>>> I think it means that relational algebra operators are not allowed to 
>>> decompose it.
>> Actually, the structure is illusory and representation-dependent. 
>> Domains have operations that appear to reveal internal structure even 
>> when that internal structure may not physically exist.

> Okay, maybe it's clearer to say that relational algebra operators don't
> decompose (attribute) values.

I did not mean to imply I disagreed with you. I don't see David's posts directly, and I responded to what he wrote because you excerpted it.

It's true the relational operations do not decompose attribute values. There are a number of contexts where one can apply operations to values including operations that seem to expose some structure.

My point was even when operations seem to expose some structure, the structure does not need to physically exist anywhere. If we are "seeing internal structure" as David says, that structure may be a figment of our imagination.

He seemed to suggest we need to have some formal way to see physical structure. I disagree with that wholeheartedly.

> (There might be operators that do decompose, eg., aggregate operators,
> but they aren't part of the essential algebra. Also, I think
> decomposing a representation structure is quite different from
> decomposing a value. I was trying to say that I think the relops don't
> do either.)

The relops work on logical structures called relations and on no other structures. Under the covers, the dbms may operate on all sorts of physical structures: pointer chains, indexes, partial aggregates, etc. Received on Wed Oct 31 2007 - 18:07:18 CDT

Original text of this message