W3C home > Mailing lists > Public > public-data-shapes-wg@w3.org > January 2015

Re: Shapes vs Classes (in LDOM)

From: Holger Knublauch <holger@topquadrant.com>
Date: Fri, 30 Jan 2015 11:24:05 +1000
Message-ID: <54CADD35.2040708@topquadrant.com>
To: public-data-shapes-wg@w3.org
On 1/30/2015 2:53, Arthur Ryman wrote:
> I disagree that "shape" is essentially the same thing as "class". I 
> can see your point if we are talking about OO data types. However, the 
> situation with RDF is very different. 

In the context of existing RDF technology, they can indeed be regarded 
as conceptually different: for better or worse the term "class" is 
already used by the RDF community and has some meaning there. So maybe 
this conflict of terminology is the main issue. Whether you use 
ldom:Shape or rdfs:Class and ldom:shape or rdf:type is from an 
algorithm's point of view really just another URI. In fact I just 
slightly generalized the LDOM algorithm description here:


and it works exactly the same way, no matter what URIs you use for those 
roles. This tells me that our seemingly different worlds are in fact 
almost identical.

> In RDF, "class" is used to classify resources. Given a resource, it's 
> either in the class or not in the class. So a class defines a set of 
> resources. The classification may be explicit via the presence of an 
> rdf:type triple, or it may be inferred via RDFS or OWL. In contrast, a 
> "shape" describes constraints on RDF graphs, i.e. sets of triples. 
> There is no requirement that an RDF graph contain rdf:type triples. It 
> may contain zero or more. In practice, many RDF graphs do arise from 
> OO data types. That is an important pattern, so we should make it easy 
> to describe. However, we shouldn't exclude the general case which is 
> also valuable and is one of the motives for using RDF instead of 
> alternatives.

I believe we need both - the ability to create independent shapes and OO 
patterns. One main motivation here is backwards compatibility so that 
there is an organic way for people to reuse their instance data without 
being forced to use both rdf:type and ldom:shape/oslc:instanceShape. 
There are many other reasons. You state elsewhere that you would have 
used OWL Closed World, so you don't seem to have had issues with reusing 
the concept of classes in the past. If we can agree on that middle 
ground, then I think we have a productive way forward.

Received on Friday, 30 January 2015 01:24:41 UTC

This archive was generated by hypermail 2.3.1 : Friday, 30 January 2015 01:24:41 UTC