Re: Scope of blank node labels in TriG/N-Quads

On 6 Sep 2012, at 04:21, Sandro Hawke wrote:
>> First, backups/restores of graph stores. This can be addressed using skolem IRIs.
> What if the graph store already has both Skolem IRI nodes (acquired from Web crawling) and blank nodes?   How can you make the backup/restore faithful?

Systems may wish to mint Skolem IRIs in such a way that they can recognize the IRIs as having been introduced solely to replace a blank node, and map back to the source blank node where possible.

IOW, a store can mint the skolems in a way that allows it to tell apart its own skolems from crawled skolems. That's easy enough to do. I believe Steve said that 4store or 5store already implement it this way.

>> Second, separation of inference [1]. This doesn't require sharing of blank nodes in surface syntaxes, but only *within* a store, so it's not relevant here.
> Why is that only within a store?

It's what [1] describes.

> Alice has some data, does some inference, tracks that in some provenance, gives the resulting dataset to Bob for display/additional-processing/analysis.     Alice gives the data to Bob as a TriG file.

Well but that's a different use case with different considerations. In [1] it was about being able to update the inferences in response to changes in the base data. That doesn't work here because Bob can't match the blank nodes in version 2 of the TriG file back to those mentioned in version 1.

So I don't see why Alice needs to keep the inferences separate -- she can just stick inference + base into a single graph and say with provenance that it was derived from the other graph via inferencing.

What you're asking now is essentially the old “how do I link to a blank node” problem. The answer is, you can't do that. You can use URIs, and skolem URIs in particular.


Received on Thursday, 6 September 2012 13:37:03 UTC