Microdata, RDFa or JSON-LD Appropriate or best usage?

Question

I have been wondering which of those formats is \"best\"? Schema.org, Microdata, and RDFa are bit of a pain to implement. They can break validation and require quite an effo

Answer 1

(updating answers!)

About "popularity", please see this question/answers.

Microdata today is the most popular: in a universe of 34 million of domains, 5.63 million (~17%) use "content markup" (I will use the jargon markup) by RDFa (0,9 million), Microdata (2.5 million) or Microformats, and less than half use separated semantic descriptors, noticing the most popular as JSON-LD, with 2.12 million (6%).
PS: we prefer "per-domain statistics" (instead per-page statistics) because pages in same domain in general have same templates and other local-authority convention enforcements.

In a universe of "domains expressing semantics" (7,75 million) the statistic profile is:

• 73% markup semantic
• 27% separated semantic
• (... intersection as mix "separated+markup" can be zero to simplify...)

Rule of thumb in 2017

Use markup semantic with Microdata and, after it, if you need to express something more to machines, use JSON-LD.

---

Use markup semantic because it is the most popular, and because marked contented will be verificable/auditable simultaneously by humans and machines.

Important: remember that Microdata, RDFa (a W3C standard) and JSON-LD (a W3C standard) can be (easily) translated to RDF, so all these formats are compatible.

---

PS: for HTML tables see also W3C's tabular-metadata. For open non-HTML resources, as CSV files, use RDF-compatible W3C's tabular-data-model and/or frictionlessdata/specs.