comments on the multiple body scenarios wiki page

Tim,

I am going through the wiki page on the various options (it is a great page!). I have some comments but, instead of changing the page directly, I prefer to comment… (yeah, one would want to use an annotation:-)

- Scenario 1, all the alternatives beyond the first one: the text "This image is worth viewing on my desktop." appears as a body in the first one, but, starting from the second one, you add the english language tag. This may sound as if for all other cases the appearance of the language tag is a requirement, which is not; it is a SHOULD in the model, and not a MUST. Actually, the same holds for the format statement; that is also a SHOULD (And let us make a mental note whether that is really necessary in the specification. We are talking about a class whose goal is to include a text, the format statement seems to be a completely superfluous entry in my view. I will create an issue.). The same issue appears in the other scenarios and, obviously, is more a matter on how we define and use the model...

- The Arc de Triomphe Wikipedia URL is set as a semantic tag in the first alternative in Scenario 1, but that is turned into a simple URI in the second. Isn't that a mistake?

- In alternative 3 of Scenario 1, shouldn't we have a 'role=tagging' attribute added to the wikipedia reference? I realize that the SemanticTag class is around but, in this scenario, it is actually fairly strange to have role in some cases and not in another. That may reveal an ambiguity of the model…

- A general comment on the Turtle encodings: the turtle side may actually be simpler; in many cases, where you did use of the {…} syntax in JSON-LD to create blank nodes, you did not use the corresponding […] syntax in the turtle side. Of course, the issue here is not to simplify turtle, so it is not that important:-)

But these are all details…

There is, however, a dangerous semantic bug that may be created through alternative four ("Role as Class / Typed Bodies and Targets") which is probably worth emphasizing in the text (my apologies if this issue was discussed during my vacations). If I take a simple setup from the Model spec:

{
  "body": {
      "id":"http://example.org/body1",
      "type":"dctypes:Sound"
   },
   "target" {…}
}

And I want to add a role to the body, I may be tempted to do

{
  "body": {
      "id":"http://example.org/body1",
      "type":["dctypes:Sound","commenting"]
   },
   "target" {…}
}

which yields the triple

<http://example.org/body1> rdf:type "oa:commenting"

which has exactly the same problem as the original issue that started this whole thread, namely to put a general typing onto the web without the proper context. Ie, my example above is wrong. I think that this approach may become pretty error-prone in practice. Maybe it is worth emphasizing on the page.

(You avoided that through yet another level of classing through the use of "item"…)

Ivan


----
Ivan Herman, W3C
Digital Publishing Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
ORCID ID: http://orcid.org/0000-0003-0782-2704