- From: Richard Cyganiak <richard@cyganiak.de>
- Date: Sat, 1 Oct 2011 11:57:48 +0100
- To: William Waites <ww@styx.org>
- Cc: Sandro Hawke <sandro@w3.org>, Arnaud Le Hors <lehors@us.ibm.com>, public-rdf-wg@w3.org
On 30 Sep 2011, at 11:45, William Waites wrote: > cygri> Anyone can always add another *triple*. If we want to use the > cygri> fourth element in assessments of provenance and trust, we > cygri> can't let anyone add arbitrary *quads*. > > Ok, so *where* can anyone add another triple? What exactly is the > restriction on the fourth column and how does it work? Let's distinguish between “private” datasets, and the “global” collection of graphs called the World Wide Web. In “private” datasets, you can do whatever you want, because no one can stop you from doing that anyways. You can add any quad. You can use any IRI as a graph name. In the “global” dataset, the IRI owner gets to say whether any given IRI is actually a graph name, and which triples are in that graph. IRI owners have several options for announcing which triples are in which graph. The most straightforward one is to make the graph IRI dereference to a representation that contains those triples. But other schemes are being used in practice. This means that the “global” dataset can only have “owned” IRIs as graph names. There is just no sensible context-independent way of determining which triples are in a graph named with an un-owned IRI, such as <urn:uuid:c2f41010-65b3-11d1-a29f-00aa00c14882>. Of course such IRIs can be used in “private” datasets just fine. I think the job of this group is to specify a data structure that can be used to express both the “global” and “private” datasets. I don't think we have to get into the business of explaining all the social conventions around “global” datasets. Best, Richard
Received on Saturday, 1 October 2011 10:58:14 UTC