Re: One-To-One Relationships
Date: Sun, 2 Dec 2007 06:00:29 -0800 (PST)
On 2 dec, 13:37, JOG <j..._at_cs.nott.ac.uk> wrote:
> On Dec 2, 10:46 am, Jan Hidders <hidd..._at_gmail.com> wrote:
> > On 1 dec, 01:13, JOG <j..._at_cs.nott.ac.uk> wrote:
> > > On Nov 30, 9:36 pm, Jan Hidders <hidd..._at_gmail.com> wrote:
> > > [snip]
> > > > You want me to explain the difference between the notion of "thing"
> > > > and "fact"?
> > > Yet /again/ you are put new words in my mouth. I never mentioned the
> > > word "fact" in this context. I asked you to elucidate the distinction
> > > between 'entity' and 'relationship', following your condescending
> > > statements (and *sigh*'s) that the issue has been resolved, inferring
> > > that not just me, but everyone involved in the debate here was foolish
> > > even to be discussing it.
> > Hmm, you're reading a bit more into my sighs than I intended you
> > to. :-)
> Ah the joys of non-verbal communication. Perhaps I reacted too
> vociferously, so apologies for any over-defensiveness.
> >I wasn't so much putting words into your mouth as telling you
> > what I think the definition of entity and relationship are. An entity
> > is just a thing, and a relationship is just a fact, or, to be bit more
> > precise, an instantiated predicate. Wasn't that what you asked?
> Yet I find those definitions entirely unsatisfying. Is a relationship
> not a 'thing' too?
Yes, it is.
> Its a noun after all. And how else can one possibly
> describe an entity apart from through 'facts'?
One can't. But it is not necessarily true that all these facts are in your Universe of Discourse. In logic, for example, this assumption is not made. Of course, when making a database schema, it is a very bad sign if they are not, but I wouldn't rule out the possibility in principle. And even if they are, you could still wonder if these facts are necessarily all directly related to the entity in the sense that the entity plays a role in the fact. Weak entities, for example, are a case where this is not true. So how restricted should the notion of "identifiability" be in a database schema?
> In addition the 'entity as thing' definition rather passes the buck as
> to defining what a 'thing' is. For a field exhibiting the rigour that
> mathematics does it all seems very hand wavy and fait accompli. It is
> possible that Deutch's logic of relative identity is attempting to
> solve all those identity puzzles after the horse has already bolted,
> and that there is something incomplete in our notion of entities that
> is generating the paradoxes. I just have this nagging feeling that
> there is something more subtle going on.
Sure. There actually have been papers in the database community on formalization of role-dependent identity,e.g., http://citeseer.ist.psu.edu/32127.html but IMO it never really got the attention it deserved.
> [...] My impression is that
> proponents of OO and E/R as a model unto itself, promote entities as
> first class citizens of a model, whereas RM proponents and logicians
> do not, utilizing them as second class citizens to be talked /about/.
If with "promote entities as first class citizens" you also mean "an thereby demote relationship as second class citizens" I'm really puzzled where you get this impression from. That is not my opinion, and as far as I know them also not the opinion of those that promote (some dialect of) the ER model as a model unto itself (that would be mostly the ORM crowd). So I wonder where you got that impression.
> Absolutely not. At the conceptual level I believe this is how most
> people think, and the model is their to facilitate them after all. I
> just believe there are numerous conceptual models, all equally valid,
> that can emerge from a neutral logical one, and I fear entity-
> relationship thought as an end to itself can get one stuck in one
> brittle viewpoint (this is the same concern I have with XML).
Good, then we have still something to disagree about. :-) My claim is that a properly defined ER model would actually be better as such than the RM.
- Jan Hidders