Example 3 is misleading v.a.v. JSON-LD

[iherman]

The text before Example 5 gives the impression that example 5 is, in fact, in JSON-LD. Unfortunately, this is not true, because the @dir is not a valid keywords in JSON-LD 1.0 and is currently not included JSON-LD 1.1.

I have just raised an issue for JSON-LD 1.1 to include a separate @dir keyword albeit without mapping into RDF that cannot (alas!) properly represent the direction of a string literal. As W3C is considering the creation of a JSON-LD Working Group, that may eventually officialize JSON-LD 1.1, this may at least settle this problem on the JSON-LD level.

[aphillips]

I've updated the text here (the document itself has been hugely reorganized in the past week). I agree that there are a number of potential action items. This issue being kept open because not yet ready to close.

[iherman]

@aphillips, note that, since then, the relevant JSON-LD issue has been closed with no action. The overall feeling is that the missing directionality is a matter to be solved on the RDF level and adding some hack into what is, after all, a serialization of RDF is a mistake.

[iherman]

@aphillips I looked at the latest document (dated February 13), and the example is still wrong. There are several issues.

(The example number has changed. It is now Example 3, in §5.2. I have changed the title of the issue.)

• (minor) context is incomplete; it should at least include a @vocab to make it correct.
• it is unclear whether the intention is to use value, language, and dir are meant to be aliases of the corresponding JSON-LD keywords (for the first two) or whether they are 'just' JSON-LD terms.
  • In the former case (see JSON-LD playground example) the JSON-LD is simply incorrect; the dir argument is not accepted per JSON-LD syntax
  • in the latter case (see JSON-LD playground example) what you get is not what you expect. The value of "title" is not a string, but it is a separate object with three values which, from a JSON-LD data model point of view is fundamentally different.

Similar errors appear in Examples 4-7, which make them all invalid.

(This is also one reason why Example 4 does not work, ref #13.)

My overall advise would be not to use JSON-LD in the examples. On the other hand, it should be noted that JSON-LD is constrained in its syntax by the current restrictions of RDF and, thereby, only recommendation in §4.5 seems to be feasible for JSON-LD based encodings for now as far the base direction issue is concerned. Specific applications of JSON-LD may also set a default value for the base direction but... that is all that can be done for the time being.