Re: Database design, Keys and some other things
Date: Wed, 28 Sep 2005 22:06:16 +0200
Message-ID: <433af79b$0$11064$e4fe514c_at_news.xs4all.nl>
JOG wrote:
> vc wrote:
>>JOG wrote: >>>vc wrote: >>> >>>>'The sky is blue in the daytime' ain't no predicate. It's an [ambiguous] >>>>proposition which could be false or true if it were not ambiguous. >>> >>>Ok. There exists a meteorlogical observation where [the sky is blue in >>>the daytime]. That is the sentence (not a very good one i'll agree, but >>>a propositional sentence nonetheless), with the predicate in square >>>brackets. >> >>Still not good enough. In logic, a predicate is a statement whose >>truth depends on the variable(s)ranging over some domains(s). E.g 'x >>is_older_than y' is a predicate, but 'John is older than Jim' is a >>proposition. You need to restructure your sentence some more in order >>to make a predicate out of it.
>
> Ok, thank you for the correction. I'm using weak terminology in a very
> specific area (using predicate in terms of its grammatical meaning
> instead of math logic).
>
>>>P = { <feature: sky>, <colour: blue>, <period: daytime> }
>>>
>>>M = { <creator: James>, <created: 1127871055>, <statement: P> }
>>
>>OK. All the talk about predicates aside, why the piece of information
>>above should be treated in a special way and called by the nebulous
>>word 'metadata' ?
>
> Yes talk of the terminology is a side issue. First I agree the term
> 'metadata' is woefully blurry, and given all data may be metadata and
> vice versa, its somewhat of a meaningless concept.
It boils down to "what do we need to know?" The topic of the propositions doesn't matter in a way that would affect the way we design a database. Data about data - metadata - so what?
Databases and databased thinking help when
there are a lot of propositions to just a
few predicates. It breaks down when we want
to have propositions, not knowing the
predicates in advance.
>>What is gained by this in comparison to treating the >>additional attributes as part of the original entity (or being a >>separate entity) ?
>
> Great question, and not one I'm sure I can answer yet - outside the
> fact that its mathematically correct to do so. Storing data _about_ a
> proposition in with the encoding of the proposition itself seems to me
> like attributing a quote to an author by putting his name inside the
> quote marks.
So you want to know something about the propositions. Great! Tell me exacly what it is you want to know about them and together we can design the database just for you. I'll repeat: the topic does not matter.
You are not the first who wants data about data, facts about facts. Disasters like Enron's caused a lot of people to seriously rethink about what we need to know about what we know, and Larry E was rather fast to respond with several features in Oracle 9 & 10 to cater for frequent needs.
> I mean, RM's still do a grand job, but given part of their allure is
> their theoretical correctness, it seems to me worthy of checking out as
> an area of possible improvement.
>
Received on Wed Sep 28 2005 - 22:06:16 CEST