Re: Does Codd's view of a relational database differ from that ofDate&Darwin?[M.Gittens]

Alexandr Savinov wrote:
> ...
> Assume you have two tables Products =<Price, Size> and Personel = <Age,
> Salary>. Then you have a record in Products r=<$100, big>.
>
> Question: What Age has the record r?
>
> Answer: null.
>
> ...

i admit that i don't understand this thread but for some reason i stumbled over part of it.

i presume the question above can be restated as "how old is the big $100 product?". if so does the 'null' answer above mean "inapplicable"?

if the question means something a little different, say "what are the salaries of the Personnel who might have sold the 'r' product?", would the answer (for whatever it's worth) not be the set of salaries of present personnal? (because <AND>ing the two 'tables' ala TTM, gives their cartesian product or to say it another way because there is still a common attribute namely the empty set). no matter whether i project first or last, i seem to get essentially the same answer.

my cut at all this is that "how old is the big $100 product" is an example of a kind of question that can't be answered by the RM, given this schema. the "what are the salaries ..." question might be pointless in practice. it does seem to be of the kind that can be answered but i'm not sure if this is always so. it seems more precise to say that the 'null' answer is really not an answer to the original question, rather it is an answer of sorts to the conditional question "can the original question be answered?" (which has only two possible answers - true and false. while i can imagine a null advocate saying that the answer would be 'unknown' were it posed to my landlady's dog by an optimist, i'd say that in any real situation the answer is only true if a set of tuples, empty or not, can be produced in reply to the question).

i suppose my question comes down to this: is it useful to limit RM use to questions of the second kind? ie. do we really need to answer any other questions besides ones that are patently answerable with a set of tuples? or to put it more simply, do we need any answers that aren't sets of tuples?

just puzzling out loud,
p Received on Fri Jun 10 2005 - 23:11:59 CEST