RE: Generated RDF conformant with good practise?

> The question of whether there *is* type info present in the original XML or
> not is one that endlessly fascinates (as you can see, it doesn't take much
> to amuse me :-). I think you could parse it either way: even without a
> schema present, <bib> could be considered both a type and an instance. Or
> does madness that way lie? :-)
> 

That's exactly right.  In essence a specific xml element is an instance of a type (or "class", depending on your use of terms).  The type is named by the element name.  Of course, for some types of markup, this notion is not useful, as for <strong>inline</strong> elements.  But when the markup attempts to describe "things", it is pretty applicable.

[In a different post]
>  An author is an instance, a real thing. How is a bib having books 
> in some substantive way different from a bookhaving authors? 
> Why do they need different relationships to describe them?

As I said, it is a  modeling issue, or as you said, a conceptual one (the two are often very much the same kind of thing).  There is a relationship between a book and its author.  You want to describe (some things about) the book, and you need to identify the kind of relationship.  Sometimes, this particular kind of relationship is called "author".  Note that a person plays the author role in contrast to an author being a subclass of Person.  Thus it makes sense (to me, anyway) to call the relationship by the term "author".

In the case of the bibliography, a book does not play a role in the bibliography, not in the way that author plays a role in a book.  So I don't favor naming the relationship between bib and book by the term "book", though of course you can call it anything you like.  

Actually, I would say that a bibliography consists of "entries", where each entry describes a book or other publication.  The relation between a bib and its entries could well be called "entry" or "bib-entry".  The resource  representing an "entry" might be called a class "BibEntry".  Now a bibEntry could well have a property called "book".

Again, it's a modeling matter.  It doesn't even have much to do with RDF specifically.  You would go through the same exercise if you were going to model the thing in a relational database.  XML lets you avoid some of that if you want - the missing parts are implied (to you, anyway, and hopefully to others), or you don't consider them important to spell out, such as the relation between a bib and its entries, because you think they are obvious or don't need saying.

In this kind of modeling, there is often no one right way to do it.  The answer may depend on how you see the structure of things, and what you want to do with the modeling information.

Cheers,

Tom P

Likewise, there

Received on Thursday, 16 September 2004 17:31:31 UTC