Re: [TF-TD] experimenting with mapping JSON based data models to RDF triples

> On 28 Dec 2015, at 16:34, Melvin Carvalho <melvincarvalho@gmail.com> wrote:
> 
> As an independent web developer I personally find it easier to work with RDF triples than with JSON.  One single advantage of JSON imho is that it does not require a parser when working with JS.  But adding a parser is one like of code for me, so not a huge overhead.  But I suppose everyone will have their own preferences.  So thanks for putting this together!

Some further thoughts:

A direct representation of raw triples would be very awkward as an authoring format for data models, and likewise for visually inspecting the data models exposed by a thing. The history of the Web shows that this matters, e.g. the dislike for XML namespaces.

A binary representation of triples assuming shared dictionaries would enable compact encodings.  For instance, you could encode URIs as small integers that reference the shared dictionary.

Object data models scope property names to the object they are declared for.  When it comes to mapping to RDF, we should allow for the re-use of names in different contexts, e.g. for properties and for action parameters. One way to deal with that is to use the path from the root of the data model as a prefix that disambiguates the leaf names. This makes triples even less appealing for authoring.

In essence, RDF is generally a good fit for metadata, but less so for object data models.

—
   Dave Raggett <dsr@w3.org <mailto:dsr@w3.org>>