Re: “RDF systems” and empty graphs (was: Re: Making progress on graphs)


On 15 May 2012, at 15:53, Sandro Hawke wrote:
> Given this framework (which seems quite reasonable), it seems clear to
> me that SPARQL Update 1.1, as a spec, would not conform to RDF, if RDF
> Concepts defined Dataset as you propose. 

How could this possibly be the case? SPARQL 1.1 Update is defined in terms of “graph stores”. Graph stores allow empty graphs, exactly like my proposal. If it contained a bit of additional language that related the term “graph store” to the term “RDF dataset” or whatever we end up calling it in RDF Concepts, then

> I mean: wouldn't RDF Concepts be setting the expectation that specs
> using "RDF datasets" would mandate systems support all the features of
> datasets?

Well, if conformance is properly and clearly defined (which it may not yet be in RDF Concepts), then we don't need to argue about the expectations that are set, but simply need to read what is required to conform.

RDF Concepts ED currently says:

Another specification conforms to RDF 1.1 Concepts and Abstract Syntax if it defines operations in terms of RDF graphs or RDF datasets, and if any use of terminology defined in normative sections of RDF 1.1 Concepts and Abstract Syntax is consistent with its definitions and conformance requirements.

This is meant to say that another spec simply has to import the definitions from RDF Concepts and use them without contradicting itself. And define some behaviour that involves RDF graphs or RDF datasets. That's all.

So, a hypothetical spec that says: “This is a syntax for serializing the subset of RDF datasets that contain exactly three named graphs and one occurrence of the letter E” conforms to RDF Concepts.

> For clarity on this conformance perspective, I don't think the RDF/XML
> spec will conform to RDF Concepts going forward.  

You mean because it cannot serialize all graphs? As long as the spec doesn't claim to be able to serialize *all* RDF graphs, and especially if it only claims to serialize a certain well-defined subset of RDF graphs, then it should be considered as conforming to RDF Concepts.

Consider the RDB2RDF Direct Mapping for example. It maps databases to RDF. So its output are RDF graphs, but it cannot create arbitrary RDF graphs. Nevertheless I don't think there's a possible coherent argument that it doesn't conform to RDF Concepts.

So I consider SPARQL Update as conforming to a definition of datasets that allows empty graphs, even if conforming SPARQL Update implementations are allowed to drop such graphs.


> I assume we're
> granting it an exemption for historical reasons, but I think we would be
> quite resistant to any new RDF syntax had those same limitations.
>   -- Sandro

Received on Tuesday, 15 May 2012 20:41:57 UTC