Re: why I don't like named graph IRIs in the DATASET proposal

Richard,

On 10/01/2011 02:22 PM, Richard Cyganiak wrote:
> The triple
>    <http://example.org/bob>    dc:publisher  "Bob" .
> provides information about an entity of unspecified
> nature. It says that this entity was published by Bob. The entity is
> known to be denoted by an IRI that is also used as a graph name in
> the dataset. As a convention it seems quite reasonable and useful to
> make the assumption that the mysterious entity denoted by the IRI is
> actually the graph.

As a convention, it *does* seem quite reasonable and useful to make that
assumption.

However, if it is not *required* by the dataset definition, that
assumption may prove wrong in some cases (as reasonable and useful as it
may be), so without any additional knowledge, one should refrain to
infer anything from it.

That, in a nutshell, is what bothers me with the way the dataset
proposal defines graph IRIs. I would feel much more comfortable if it
*did* require that the graph IRI denote the graph (g-snap or g-box, that
would still have to be discussed), and promote your assumption to a
necessity, so that one can *safely* rely on it.

Granted, anyone can still populate their SPARQL datasets with ill-chosen
graph IRIs, just as anyone can use RDF without as a graph data
structure, without caring about its semantics... There is nothing we can
do about it, but I don't think it is a reason to give up on useful
semantics completely.

> But I think we can leave this association in the same nebulous realm
> of convention and social contracts as the association between IRIs and
> their referents in RDF in general.

As you stress it, RDF does not dictate which IRI should denote which
resource (including graphs). I don't think I ever suggested to change that.

However, RDF dictates that each time I use the same IRI (as a node), it
denotes the same resource.

What I suggest is to extend the above principle to: each time I use the
same IRI (as a node *or a graph name*), it denotes the same resource.

As you point out, this seems reasonable and useful, all the more that in
SPARQL or Trig, the IRI as a node and the IRI as a graph name are
represented by the *very same* lexical token (the IRI betwen <>
brackets). So my bet is that so many people will (or already do) take
this assumption for granted that it even seems unwise to allow it to be
unmet.

 best

  pa

Received on Sunday, 2 October 2011 17:07:25 UTC