Re: Interesting article: In the Beginning: An RDBMS history
Date: Wed, 05 Apr 2006 15:34:31 GMT
Message-ID: <bsRYf.225411$B94.102419_at_pd7tw3no>
x wrote:
> "paul c" <toledobythesea_at_oohay.ac> wrote in message
> news:xPbYf.218029$B94.60669_at_pd7tw3no...
>
>>x wrote: >> >>>"paul c" <toledobythesea_at_oohay.ac> wrote in message >>>news:GX_Wf.201803$sa3.143853_at_pd7tw1no... >>>
>
>
>>Thanks for that, anyway. Maybe I was dreaming about what I thought I >>read. I see the word 'relation' doesn't appear in the quotes but maybe >>this doesn't matter (assuming that those drafts' other words appear in >>the actual standard with the same intended meanings).
>
>
> I think Jonathan Leffler was right when he mentioned the word
> "relationships" may refer to something else.
> For example I've seen it used for <, >, >=, = among other things :-)
>
Being as thorough as he usually is, I'm sure Jonathan's right too. I'm darned if I know what a "relationship between tables" is unless it's another table. For that matter, I don't know what the sql standard would mean by "table" (assuming it uses that word). I've assumed that it doesn't stand for a relation partly because it allows duplicates and nulls. Without those differences, I imagine an sql table still couldn't stand for any relation we choose because at least when I was using it ten or more years ago a row-column intersection contained only a single value, ie. some relations can't be expressed as one table.
>
>>Codd, in his 1970 paper, pg 80, said: "Accordingly, we propose that >>users deal, not with relations which are domain-ordered, but with >>relationships which are their domain-unordered counterparts". In the >>next paragraph he says: "To sum up, it is proposed that most users >>should interact with a relational model of the data consisting of a >>collection of time-varying relationships (rather than relations". >>It seems that by using the word "relationship" he was merely trying to >>draw a line between mathematical relations that have ordered domains and >>ones that replace the ordering with names as well as between logical and >>physical representations. (By "most users" in the second sentence I'm >>guessing that he was talking about everybody except DBA's.) Also seems >>that what he meant by 'relationship' was roughly what we call 'relation' >>today.
>
>
> About "domain ordered".
>
> What do you think is the difference between the Codd pi
> (projection/permutation) operator and Chris Date rename coupled with
> projection operators ?
>
> To me they appear to be the same.
>
> For example pi21(supply{project, supplier}) appear to be
> select project as supplier, supplier as project from supply
>
>
>
>
>>So I guess it would be reasonable for the standard to use "relationship" >>as long as it clarifies where it uses that word in some special sense, >>such as when talking about functional dependencies which seems to me to >> throw 'relationship', 'table' and 'columns' up in the air and let them >>land on the floor however they may. That heading 4.14.4 about >>"Relationships between tables" seems murky since Codd's 1970 paper >>AFAICT mentions tables only in the sense of 'data description tables' >>(presumably in the dictionaries of the hierarchical products of those >>times, the 1969 paper doesn't seem to mention them at all).
>
>
> They use "relationships" for a lot of things.
> It appear they use the relational model to implement the SQL but they don't
> allow the users to take advantage of it.
>
>
"They" included some smart people so that could well have been in the back of their minds but if the standard doesn't mention 'relation' it would seem that it would be incapable of saying that which I think it ought to if that's what's intended. After all, smart people aren't always right!
Anyway, Jonathan's count of 1279 pages for the 'fundamentals' seems pretty damning to me. Would mean an individual's life work to understand it unless it is organized extremely well, eg. 50 actual fundamental pages and 1229 of explanation/exposition.
p Received on Wed Apr 05 2006 - 17:34:31 CEST