- From: Peter Ansell <ansell.peter@gmail.com>
- Date: Wed, 7 Apr 2010 11:55:42 +1000
- To: Vasiliy Faronov <vfaronov@gmail.com>
- Cc: Ian Davis <lists@iandavis.com>, Leigh Dodds <leigh.dodds@talis.com>, Linking Open Data <public-lod@w3.org>
On 7 April 2010 10:31, Vasiliy Faronov <vfaronov@gmail.com> wrote: > Hi Peter, > > Thanks for your perspective--this problem is something I haven't > considered before. But let me bring up a question. > > Suppose we have the resource (URI) X, which is the original resource > that the annotations are all about. Someone comes around and publishes > an annotation about X, using a newly minted URI Y (owl:sameAs X). > > Y conforms perfectly to the Linked Data model: you can dereference Y and > get the annotation for X, then follow the owl:sameAs link and get the > "authoritative" description of X. > > But the question is: where else in the Linked Data graph would we find > references to Y? Since everyone wishing to annotate X mints their own > URIs, we end up with a number of equivalent URIs that all point to X, > but are themselves not pointed to. In other words: using URI Y in a > description would constitue an annotation, and thus, according to your > model, warrant minting a new URI Z. So, even though Y can be linked to, > nobody actually ends up doing this. > > Maybe I'm missing some other uses for Y? If there was an annotation, there should be an equivalency relationship defined to the original URI, so Y would link back to X, and Z would link back to Y, and possibly X. To review: * The major reason for not wanting to do this is that it requires reasoning processes to infer the relationship between Z and both Y and X. * The major reasoning for wanting to do it is that non-authoritative sources can be referred to, and the references can be resolved using Linked Data principles, without requiring the original authority to participate, as they may not want to do so for any number of social or technical reasons. * My contention is that it is not a good idea to promote the idea that everyone use X in their documents, if the statements are not going to be resolvable using the URI's in the statement, as they would be if people used Y and Z. I guess it is a fundamental mistrust in the social aspects of the Linked Data X-only system that requires everyone to accept everyone elses ideas. This mistrust does however have practical benefits, as it results in a system where statements are resolvable using information in the statements, so there are social benefits to the system. In my opinion, the benefit of creating Y is to provide accessibility while we are still emphasising direct HTTP GET resolution above other future recommendations such as pinging multiple SPARQL endpoints for information in order to resolve a URI. If distributed querying was common than a sole URI, Y, would be useful, as there would not be a simple one to one restriction between Y and *the* authoritative document describing Y because a user could compile their own document using their trusted endpoints (or simply all endpoints if they have no trust opinions). Cheers, Peter
Received on Wednesday, 7 April 2010 01:56:15 UTC