Re: RDF/JSON-LD representation of context

Hi Lukas,

I must admit that such thoughts have occurred to me (on more than one
occasion), but I feel a warning is in order. This would be roughly
equivalent to represent e.g. Turtle prefix declarations in RDF, and further
down the line, maybe "reifying" other kinds of structural details into RDF
itself. (Actually, some of that has been done. See e.g. VANN [1].) This
crosses the representation/expression barrier that is so useful to
insulate, since it leads to complexities of self-reference and quoting.
Also, should clients support both "encoded" and "expressed" context
definitions (I'd say no)?

It would benefit to clearly explain why this is actually needed, otherwise
it might become a rather abstract exercise. I've had one case of late
though: defining RDF processing pipelines controlled by rules given in RDF.
I haven't yet found this compelling enough to warrant more work, but if
that's your use case I'd be interested in the details.

This is actually tangential to the mechanics of CSVW. You may want to look
into that too.

Cheers,
Niklas

[1]: http://vocab.org/vann/#preferredNamespacePrefix


On Wed, May 10, 2017 at 12:04 PM, Lukas Rosenstock <lukas@cloudobjects.io>
wrote:

> Hello everyone!
>
> I’m curious if there’s any existing work on specifying a JSON-LD context
> as JSON-LD itself or any other RDF representation; in other words to
> express the term definitions as Linked Data.
>
> In a system in which all data is stored as RDF and an API provides access
> through different RDF serializations including JSON-LD it could be helpful
> to store and express a context in the same format as the data itself.
>
> For example this context
>
> {
>     "name" : "http://example.com/name"
> }
>
> could look like this in RDF
>
> {
>     "@id" : "urn:samplecontext",
>     "jsonld:definesTerm" : {
>         "@type" : "jsonld:TermDefinition"
>         "jsonld:hasKey" : "name",
>         "jsonld:hasUri" : { "@id" : "http://example.com/name" }
>     }
> }
>
> Is someone doing work on this or, if not, would there be anyone interested
> in this, too?
>
> Regards,
>
> Lukas Rosenstock
> https://cloudobjects.io/
>
>
>
>