Making CSL/JSON the default interface for expressing references in Texture

[michael]

We use an internal Substance model during editing since we don't want to have nested JSON structures. However since we have an API abstraction now we are thinking of introducing the following methods:

api.getReference('r1') // returning a CSL/JSON record
api.addReference('r2', {...})  // taking a CSL/JSON record

The advantage will be that we build on an existing standard without introducing a new JSON schema. Of course for serialization we will use the JATS element-citation tag, but reflecting the same fields.

Some open questions:

• Are there any current limits to CSL/JSON that don't fulfill publishers requirements (I'm thinking of annotated titles, or fields that are not covered)
• How about math inside of reference titles (I think BibTeX can do that, but CSL/JSON?)
  • Is this a requirement for our current journals (we don't have math journals atm afaik)

@zuphilip does CSL/JSON have an answer for math and more complex annotated titles etc.? Would you be able to help us with that transition to make Texture speak CSL/JSON fluently? :)

@Melissa37 do you think CSL/JSON have all fields we need? We need to compare this with: https://github.com/substance/dar/blob/master/DarArticle.md#ref

[Melissa37]

I am sorry and a bit out of understanding this model, so I will just ask some questions that I think are of relevance to this:

• Can the JSON manage attributes on XML that are not necessarily linked to display? My examples relate to this new JATS4R recommendation that is being finalised right now: https://docs.google.com/document/d/16h5JlmhLdl-Qdx-7CPa8WPVcQZLFtfG0KdWwbwFIbnA/edit#heading=h.eeew73cotjfj

• To check references we'll all need to build automated processes on top (such as PubMed validation and also Crossref validation. Also, we have built schematron rules)...how will it work that these processes can get JATS XML and not JSON?

Cheers

[michael]

@Melissa37 you can view a reference just as an object with properties, each consisting of a value. JATS or CSL/JSON are just serialisations. Of course if CSL/JSON is missing a property we then need to introduce it. Wether a property is displayed or not is an independent story. That's the rendering part.

About PubMed/Crossref validation we should start a separate thread on. I don't yet understand how this works and what is checked.