Re: Provenance as a first-class citizen

> Sandro et. al. -
>     Perhaps I'm missing something here (I admit to not having time to
> have read the latest developments in RDF in HTML, so perhaps a quick
> refrerher is in need for the general audience), but how exactly do you
> "point to the original"? Are you saying that every RDF statement I post
> on the Web has a URI, that of the site it's posted to?
> 
> So I post
> 
>    myns:tom myns:likes myns:mary to www.example.org.
> 
> Are you saying "point to www.example.org?" to reference that URI?
> 
> What if I post two statements to www.example.org:
> 
> myns:tom myns:likes myns:mary.
> myns:john myns:likes myns:mary.
> 
> And I only want to point to the second tuple? Surely pointing
> example.org then gets us two statements, when we only want to point to
> the second? Perhaps we could use fragids to give them each URIs?  Yet
> that seems too messy...

It does get kind of messy here.  I can republish the second triple at
some other web location, and then talk about it there.  (This probably
gets more challenging if you're not using an RDF library which lets you
easily expose whatever you want on the web, running in a process which
has a stable web location.)

> Are you saying that the URI you got a statement from has the implicit
> provenance of that URI (which does make sense)?  Isn't this just equal
> to named graphs,where the name URI defaults to whatever URI you got the
> statement from? I think that sounds rather sensible.
> 
> So myns:john myns:likes myns:mary is equal to :example {myns:john
> myns:likes myns:mary.} by default.
> 
> Regardless, I think what Sandro is saying makes sense, and basically
> specifies the default behavior of quads. So, the question is whether or
> not people will pass quads around the Web (like RSS ala Ben) or just
> post RDF to URIs (via Cuire, GRDDL, etc. ala Sandro). Well, that's an
> empirical question, but obviously both are important and not
> contradictory use cases. We'll only know the empirical truth once the
> SemWeb is massively deployed. Yet on the surface these approaches seem
> complementary to me unless I'm missing something!
>                                                                        
>                                              -harry
> 

I don't think you're missing anything.

I'm all for being able to pass around named graphs, but I don't find the
lack of a mechanism for doing so to be a big obstacle to RDF deployment.

   -- sandro

Received on Friday, 17 March 2006 23:03:06 UTC