[openbiblio-dev] Multilingual matters in BibJSON

[Tom Morris]

On Thu, Apr 5, 2012 at 6:25 AM, Mark MacGillivray
<mark.macgillivray at okfn.org> wrote:
> I was
> hoping to hear back from the JSON-LD list with some feedback for our group
> before recommending, but I have not had a response yet. Are you able to give
> us confirmation that the JSON-LD approach would be suitable to solve
> language issues for your use cases?

I'm not sure how many queries you sent, but I saw this reply on the
public JSON-LD list to your initial questions:

---------- Forwarded message ----------
From: Markus Lanthaler <markus.lanthaler at gmx.net>
Date: Sat, Mar 3, 2012 at 2:19 AM
Subject: RE: query about JSON-LD
To: public-linked-json at w3.org


Hi Mark,

This is definitely the right place to ask such questions. So let me try to
answer them.


> Hence, I have now suggested to our mailing list that we adopt JSON-LD
> within BibJSON; the idea is that BibJSON remains simple, and anyone
> who wants to do validation / namespaces could do so by following the
> JSON-LD rules - e.g. identify a namespace, use it in their BibJSON,
> then write whatever tests are required within that context to check
> that BibJSON records they receive meet their needs.

I'm not really sure if I understand what you mean by validation and
especially by namespaces. Could you point me to an example how you intend to
use those namespaces? The only thing I found was
http://bibjson.org/#collection and "dc" isn't used anywhere else in that
example.


> In order to do this, I think all that is required is that we point
> people at the JSON-LD website, and do some examples; so I have a
> couple of questions:
>
> 1. the examples, I think, will just need to use the @context key
> within a JSON object to point to the relevant namespace, then use the
> relevant keys from that namespace. Am I right, or is there more?

Basically that's enough but you would to prevent to end up with just blank
nodes after a conversion to RDF e.g. you would also need to set the subject,
i.e., use "@id". You already have an "id" element but unfortunately its
value is not an IRI. Linked Data requires that you identify things by
assigning them a unique IRI.

I also couldn't find any information about the media type you are using for
BibJSON. To be Web-friendly you would need to mint a new media type for it
(regardless of whether you use JSON-LD or not). If you are just using
application/json you could link to the context by means of an HTTP header,
see

http://json-ld.org/spec/latest/json-ld-syntax/#referencing-contexts-from-jso
n-documents


> 2. Our records can come in a object under the "records" key, and that
> object may also contain a "metadata" key (hence, the total object is a
> collection). Is there a sensible way in JSON-LD to add the namespace
> information only to the metadata object, to save copying it into every
> record?

In that case you would normally add the context definition at the top-level
object. It is then valid throughout the document:

 {
   "@context": ...
   "metadata": ...
   "records": ...
 }


> 3. If you are continuing to look for uptake of JSON-LD, and if I am
> looking to recommend it to our community, would you be happy to answer
> these questions and potentially a couple more as I try to come up with
> some examples for people to follow?

I'm sure everyone of us is happy to help you. It will also help us to
understand how people are going to use JSON-LD. So, feel free to post as
many questions as you like :-)

--
Markus Lanthaler
@markuslanthaler