Re: [higgins-dev] Sub-contexts

[Sergey Lyakhov <slyakhov@xxxxxxxxxxxxxx>]

Paul,

> The RDF named graph theory makes a distinction between the main "default" graph and named, child sub-graphs. And the
> implementations e.g. NG4Java take this simple one-level hierarchy approach. 

There is no "default" graph and any graph's hierarchy in NG4J. All NamedGraphs in NamedGraphSet are independent (from each other) graphs within the same document. I supposed to use the plane NamedGraphSet in the following approach:
1.SubContext has an access to the triplets of its graph only.
2. "root" Context has an access to the triplets of all graphs in the document.

> I think we should consider support for sub-contexts to be something that CPs can optionally support. So I think we
> need some kind of API to ask a CP what its "level" of sub-context support is. [I can't remember if we have any method
> in IdAS along these lines already--indicating what "optional" features of IdAS interfaces are or are not supported by
> any given CP implementation. Sergey?]

No, we do not have such a method.

If you do mean IExtension interface added into IdAS API by Jim, 

(the following IContext methods use it: 
public Iterator getEntities(String query, Iterator attrSelectionList, IExtension[] extensions)
public String open(IAuthNMaterials authentication, IExtension[] extensions) 
)

I do not see how we can use it for SubContexts. 

Thanks,
Sergey Lyakhov

On Wed, 26 May 2010 09:25:14 -0400
Paul Trevithick <ptrevithick@xxxxxxxxx> wrote:

> Ideally we wouldn't. The problem is implementation. 
> 
> The RDF named graph theory makes a distinction between the main "default" graph and named, child sub-graphs. And the
> implementations e.g. NG4Java take this simple one-level hierarchy approach. 
> 
> So if we do decide to allow arbitrary nesting in the Context Data Model (as represented by the IdAS interfaces) we'd
> still have to deal with the fact that context providers (like the NG4Jena one) couldn't implement it. Which is
> doable--we'd just insist that CPs declare what level of support they have for this stuff. For example:
> 
> 	* support for n-levels of sub-Contexts
> 	* support for one-level of sub-Context (e.g. the NG4Java CP)
> 	* no support for sub-Contexts
> 
> I think we should consider support for sub-contexts to be something that CPs can optionally support. So I think we
> need some kind of API to ask a CP what its "level" of sub-context support is. [I can't remember if we have any method
> in IdAS along these lines already--indicating what "optional" features of IdAS interfaces are or are not supported by
> any given CP implementation. Sergey?]
> 
> On May 26, 2010, at 8:41 AM, Mike McIntosh wrote:
> 
> > I know this seems obvious to those of you that are elbow deep in this stuff but…
> > Why do we want to prevent sub-contexts of sub-contexts?
> >  
>