Re: Generalised approach to storing address details
Date: Wed, 13 Dec 2006 00:40:50 GMT
Message-ID: <m_Hfh.476881$R63.53325_at_pd7urf1no>
Neo wrote:
>>Yes, can't argue with that and I would say that my understanding is that >>the RM can happily represent ... one hierarchy at a time ...
>
>
> One couldn't agrue (very well) if RM is happy or sad when representing
> heirarchies, but one can point out when a hierarchy is represented
> unsystematically according to the Information Principle. For example,
> in the table below, user is implying john's child is jim and jim's
> parent is john.
>
> T_Hierarchy
> Parent Child
> john jim
> ...
I don't want to get into an argument about this, I just want to say that AFAIAC we cannot at all conclude that "user is implying ...". Some group of users might agree to interpret it that way but the bare three lines above imply no such thing.
My understanding is that those three lines represent a relation and may indeed represent several other equivalent relations, eg., parent_of or child_of but not necessarily ancestor_of.
In the RM, an implementation cannot even know what a (human) parent is. All a dbms can do is preserve that relation without actually knowing what it is.
What I've been puzzling about is whether there is a way to declare two relations that are somehow equivalent, eg., somehow equatable by engine, perhaps one is a view and perhaps not, say one that talks about parents and another that talks about ancestors, or as I think JOG put it one that talks about ancestor pairs and another that talks about local pairs. I'm probably cuckoo to wonder about this, for reasons of theory in math, logic and philosophy that are beyond me, but there it is.
Thanks all, for the replies.
p Received on Wed Dec 13 2006 - 01:40:50 CET