Re: A new introduction section for RDF Concepts

Hi Pat,

Thanks for the comments. Responses inline.

On 18 Nov 2011, at 19:02, Pat Hayes wrote:
> +1 to the general idea, but I am worried about making this *too* short. Some of these ideas need a little protective exposition to guard against possible misunderstandings (of which we now have a huge collection to select from.)

I agree in general, although that's also a job for the Primer. In the introduction I tried to a) informatively introduce terms that are important but would otherwise not be defined in this document (“resource”, “statement”, “denote”), and b) explain just enough that a reader might be able to acquire an informal intuition for what any given RDF graph says.

> Might be worth emphasizing what RDF isn't, eg many people (still !) think of it as a way of describing data structures (like JSON)

That's what I think, actually :-/

> "Literals denote datatyped resources such as strings, numbers, and dates." This sounds like datatyped resources are a special kind of resource, which is a common misperception. Rather, typed literals are a special way to denote.

I know what you mean, but I think this is mostly harmless because no one really uses datatypes to denote anything but strings and numbers and the like, and no one uses IRIs to denote strings and numbers. It is true that we could have datatypes that denote people and we could have IRIs that denote strings, but in practice for most people that's useless trivia and I wouldn't want to spend space in the Introduction explaining it.

I changed the passage above a bit to make the role of datatypes clearer:

“The resource denoted by an IRI is called its referent, and the resource denoted by a literal is called its value. Literals have datatypes that define the range of possible values, such as strings, numbers, and dates.”

> "The predicate itself is an IRI and denotes a relationship type, also known as a property."  
> I don't think its a relationship TYPE, just a relation. Might be worth saying it is a binary relation, in fact. 

Changed it to say “denotes a binary relation”.

> Might also be worth saying a few words about relations with more arguments, as this is a very common sticking point for many people (and continues to be: just a month ago I had to listen to an eminence explain that RDF was useless because it couldnt represent relations with more than 2 arguments.) 

I added:

“(Relations that involve more than two entities can only be indirectly expressed in RDF [SWBP-N-ARYRELATIONS].)”

But I feel that this is really something for the Primer.

> "The statement corresponding to a triple with a blank node says that something with the given relationship exists, without explicitly naming it."  This sounds like it imples the scope of the bnode is a single triple, which is unfortunate. 

Changed it to:

“Statements involving blank nodes say that something with the given relationships exists, without explicitly naming it.”

> "By social convention, the IRI owner gets to say what an IRI denotes. They do this when “minting” a new IRI." That would be great if it were true, but in fact they mostly don't do it, even then. 
> (This whole section seems to me to be very flaky. HOW does a document specify an intended referent? HOW does publishing some RDF do that? Especially when an RDF vocabulary is required to already have "clearly established referents" ? Isnt there a circularity lurking in here? )

As you know, these questions are still being debated. It is rather hard to answer them satisfactorily, and the introduction section to RDF Concepts is certainly not the place to attempt it.

My goal here is to give readers enough information that given an RDF triple, they might be able to figure out what the author of that triple wants to communicate. To be able to do that, they need to have a tiny little bit of an intuition about the way reference works on the Web. I think that the subsection as written delivers that. My goal was certainly *not* to write the definitive account on httpRange-14.

Also, the section is non-normative, clearly states that the issue is out of scope for RDF Concepts, and points to further reading.

The alternative would be to say nothing, and I don't think that that would serve readers better.

> "An RDF vocabulary is a collection of IRIs with clearly established referents " 
> Nah, there aren't any of those the be found. And AFAIK, there is no established and recognized way to clearly establish the referent of any IRI, other than http-range-14 in the 200-response case. 

Besides what I said above: There certainly are a fair number of established and recognized vocabularies that are used consistently, so even if we're not sure *how* they did it, they somehow managed to nail down their referents well enough for practical use.


Received on Sunday, 20 November 2011 20:11:11 UTC