Re: Toward easier RDF: a proposal

On 11/27/2018 5:12 AM, Holger Knublauch wrote:
> Tom certainly has a valid point, and I think it's worth drilling into 
> these distinctions to clarify where the sweet spot would be.
> 
> There are various ways that XML trees can simulate graph structures, 
> e.g. by marking certain attributes as ID or IDREF in DTDs. Likewise, for 
> JSON the native runtime structure is technically an object model that 
> can represent graphs. In both cases, these are arguably a bit artificial 
> and inconsistent while RDF is designed from the ground up to be a 
> "native" graph data model. These differences manifest themselves in how 
> the APIs look like, how the languages are explained etc. An XML user is 
> confronted with a variety of different attribute patterns on how object 
> relations are represented (even GraphML), while only the 'child-nodes' 
> relation is truly native to the APIs. For JSON, JSON-LD illustrates how 
> graphs can be represented, yet if you load a JSON-LD file, you still 
> don't get native JSON objects in their most natural form.
> 
> So let me rephrase:
> 
> XML and JSON are primarily designed for tree structures.
> RDF is designed from the ground up for graph structures.

I'd like to say that XML was primarily designed for *constructing 
languages* for data interchange (especially document-like data).  It's 
common for some XML to be thrown together to carry some data, and that 
data is usually referred to as XML, but really that kind of usage 
amounts to creating little, mostly undocumented DSLs.  XML is only a 
framework for doing that.

RDF seems to have been designed primarily for modeling or capturing 
knowledge-like data.  These are very different goals.

There are languages for RDF interchange (Turtle, RdF/XML, etc) but the 
core of RDF stands apart from them.

Received on Tuesday, 27 November 2018 15:31:46 UTC