Raw

On Using named graphs to model Accounts it states:

It is very important to recognize the following:

A graph's name does not identify its contents.

That is, two graphs with the same name does not imply that they contain the equivalent RDF subgraphs.

SPARQL 1.1 effectively supports that, saying:

the relationship between an IRI and a graph in an RDF dataset is indirect. The IRI identifies a resource, and the resource is represented by a graph (or, more precisely: by a document that serializes a graph).

On face value that would seem to undermine all potential for using named graphs for provenance (and many other things).

But Richard Cyganiac's response is on the nail:

it is only the social contracts and conventions around URI ownership and web architecture that...allow us to maintain the fiction that URIs in RDF actually identify specific entities

Right, it's axiomatic that URI owners get to decide what their URIs identify. But ok, it's hard to see how (say) my personal assertions of graph identity offer a neat mechanical solution that can work at Web scale. However the perspective set up by all this is misleading. Resources are conceptual entities: "my blog" (with the URI http://dannyayers.com/) is a resource that has representations that change fairly frequently, including a named graph. But this doesn't mean to say that it's rendered useless for provenance purposes. The claim "Danny said on his blog" makes sense even if the statement to which it's referring isn't immediately available at http://dannyayers.com/ (incidentally there is a bigger graph available). Linked data provides a means ("follow your nose") by which related information can be discovered. In fact the claim still makes sense if the original statement is no longer on the Web, it can still be useful (only the provenance paths have got a little twistier). Some data, especially if there's a way to find more data, is always more useful than none. As danbri commented:

A dc:source linking any factually-oriented document to its alleged sources goes a long way...