Re: Question on Structuring Product Attributes
Date: Thu, 3 Oct 2013 15:38:56 +0200
Message-ID: <524d7370$0$618$e4fe514c_at_dreader34.news.xs4all.nl>
On 2013-10-02 18:55:36 +0000, Nicola said:
> In article <1601143370402385874.040013hidders-gmail.com_at_news.xs4all.nl>,
> Jan Hidders <hidders_at_gmail.com> wrote:
>
>> Derek Asirvadem <derek.asirvadem_at_gmail.com> wrote:
>>> This one should not pass unnoticed.
>>>
>>> On Monday, 18 February 2013 09:09:15 UTC+11, Eric wrote:
>>>>
>>>> To quote from Codd[1970]:
>>>>
>>>> "A firstorder predicate calculus suffices
>>>> if the collection of relations is in normal form."
>>>
>>> Codd specifies what that Normal Form is, that reference is not merely a
>>> generic or general term.
>>
>> Indeed. Also note that Codd's claim is not unproblematic. Even if you
>> restrict yourself to the flat relational model, first order logic cannot
>> express many practical queries. So what exactly does he mean by
>> "sufficient" here?
>
> I interpret that quote as follows: "If we allow relations as attribute values,
> we need (at least) a second-order relational calculus". In this sense,
> first-order logic "suffices" when relations are "flat" (we don't need to
> quantify over relations). Of course I agree with you, many interesting queries
> cannot be expressed in FOL, but you can always design a query language with
> additional power (SQL is an example).
Agreed, that is indeed probably what he meant. But I guess I'm trying to argue how weak that argument actually is. Sure, keeping things first-order looks like something that keeps it simple, but does it really? If you need extensions anyway, how bad then, relatively speaking, is it to move to a higher-order setting?
- Jan Hidders