Re: Hierarchical Model and its Relevance in the Relational Model

Hi Derek,

Let's see if I can clarify my position on the questions you raise. I'm a bit short on time, so my answers will probably not be as comprehensive as they should be and will only address some of the points you made. I have papers to write and exams to be graded. :-)

Op woensdag 28 januari 2015 10:53:01 UTC+1 schreef Derek Asirvadem:
>
> The Network Model, although very relevant, is not relevant to todays issues or to the Relational Model, so let's carve that off. So we are left with the theme of the thread: you have posted, and I paraphrase, that:
>
> A. ____the Hierarchical Model rests on a theoretical void____
>
> You did not state it, but the implication is, the HM is dead, gone the way of the NM. Therefore (again, I paraphrase):
>
> B. ____the HM is dead, it has no relevance wrt the Relational Model___

You ask whether I agree with these claims. The answer is that, no, they do not correctly represent my position.

Concerning A: This is sort of true in an uninteresting way. It is indeed true that one cannot point to a single paper that defines formally the data model and the associated languages. But there are plenty of papers on defining hierarchical data models and languages to query them, and this is in fact quite a well understood area.

Some people also take "theoretical foundation" to mean that there is a philosophical foundation for that type of knowledge representation, for example such as exists for first order logic, which is more or less inherited by the Relational Model. But there are also such theories for higher-order logics, so in that sense I don't think it is true. But I actually don't accept that this is an important observation. What matter is if people in practice understand and can deal with hierarchical data. They can, and they do.

Concerning B: It becomes important here what it is that you precisely mean with "the HM". If you include the traditional assumptions about how the data is stored and the pragmatics of how to effectively query and manage it, then, yes, that is pretty much dead. But most now understand the relevance of data independence. If you define the HM as only the idea that data can be nested, then as you know some promote the idea that the Relational Model should allow this and provide means to effectively deal with them. I agree to some extent with that idea, but not when we are talking about RM as the model for the conceptual level in the DBMS.

> -- 3.1 JH Specifics --
>
> Iff you want something specific, as a starting point, try this post:
> https://groups.google.com/d/msg/comp.databases.theory/r3BHhq1EFbs/iE5AIvLD6BYJ
>
> The context of the thread, and your post itself, are not the issue, it is the statements/references/implied-statements that you made re the HM, and its value in the RM context, that I wish to take up:
>
> 4.
> > On Sunday, 21 April 2013 23:35:37 UTC+10, Jan Hidders wrote:
> > >
> > > [RM and ERM are the context]
> > >
> > > http://en.wikipedia.org/wiki/Structured-Entity-Relationship-Model
> >
> > Right. SERM. Part of Aris, a SAP thing. The hierarchical database
> > model, once again trying to make a come back in a new disguise. :-)
>
> I declare, the HM (or the "hierarchical database model") never left. It has always been here. It therefore cannot make a comeback.

Depending on what you mean exactly by the HM I would agree with that. What I was talking about above is that probably (but I might be wrong) the reason for the introduction of the notion of hierarchy is not for conceptual but for implementation and efficiency. And that idea, linking the hierarchy to how the data is stored, at least in the communities I work in, is mostly dead. Except apparently at SAP. :-)

> 5.
> > > [Rule: Circular references are not permitted]
> >
> > The German page is more explicit and does mention this rule.
> >
> > > Which appears to be a perfectly reasonable rule to me ...
> > > Is it perfectly evident that this requirement must be enforced, since a
> > > model with cyclic dependencies is plain "spaghetti", maybe even
> > > violating some normal form?
> >
> > It has nothing to do with normalization in the classical relational
> > theory sense of the word.
>
> I declare, it has a direct and definitive relationship with Normalisation, as defined in the RM.

Really? Which of the classical normal forms (1NF, 2NF, 3NF, BCNF, 4NF, 5NF [in its different variants you find in text books]) forbids cyclic dependencies?

Kind regards,

• Jan Hidders