Gregg Kellogg [email protected] On Jul 1, 2013, at 8:58 AM, David Booth <[email protected]> wrote:
> In JSON-LD, terms are converted to URIs by use of a context. However, a > context may be in a separate document that may not be accessible to a client > that is attempting to interpret that JSON-LD as RDF. Hence, the client may > be unable to determine the full URIs corresponding to the JSON-LD terms, in > order to generate the correct RDF model. Since this is likely to be a very > common problem, I think the JSON-LD spec should provide some constructive > guidance about how a client should deal with this situation. > > What might be some reasonable guidance? Something along the following lines? > > [[ > If the context for a term cannot be obtained -- perhaps because the context > document is unavailable -- then it may not be possible to reliably map that > term to the IRI that the JSON-LD author intended. In such cases, the client > interpreting the JSON-LD document MAY perform a "best guess" mapping, with > the understanding that the guess may be incorrect. Suggested "best guess" > techniques: > > 1. If a context was previously available for an version of the JSON-LD > document that is being processed, use that as the context. > > 2. Otherwise, expand the JSON-LD terms as though they are relative URIs, > relative to the document's base URI. > ]] > > Or, as a variation of #2 above perhaps a designated universal base URI > such as http://example/JSON-LD/ or http://schema.org/ . > > What do others think? This is already covered in the JSON-LD API context processing algorithm in step 3.2.3: [[[ Dereference context. If context cannot be dereferenced, a loading remote context failed error has been detected and processing is aborted. ]]] Why are we not having this discussion on public-linked-json? Gregg > Thanks, > David >
