Libero Hypermedia Walking Skeleton

Starting point

• 1 front-end app
• 1 back-end app
• Monorepo
• Run from source
• Node.js
• Bare HTML output
• Understands http://schema.org/Article and http://schema.org/name
  • Will pass through unknown properties
• Has CreateArticle and Search actions
• Ignore server-sent events
• JSON-LD only

Running it

• Run make (or make docker)
• Open http://localhost:8080 to see the public-facing app
  • Fill out the 'Create an article' form
  • See the result appear in the list and basic search
  • Look at the logs to see responses from the API (lines beginning [Axios][Response])
    • Especially actions being created, which have their own URI.
• Open http://localhost:8081 to see the API
  • See links that can be followed, and actions that can be performed

Notes

• JSON-LD will need a wrapper around it, as it can be in lots of different forms (eg single value, multiple value)
  • Doesn't appear to exist though?
  • Sci.pe have their own library, with helpers like get-value()
• No tooling available about using Schema.org as hypermedia
  • Got me thinking about Hydra again, which does having tooling in JS (eg Alcaeus)
    • Still hard to understand, and not sure it adds anything over Schema.org
• Would Schema.org really be better than going for more specialist vocabularies (eg the SPAR Ontologies)?
  • Feels easier to adopt (probably a lot more familiar), but won't cover everything that we need
• Started to look at rdflib.js
  • JSON-LD support appears to be broken, Turtle worked
  • API is quite nice, but going all in on RDF is probably a set to far
  • Hard to see how it would be persisted without using something completely unfamiliar (SPARQL, supported by Amazon Neptune)
  • Got Tim Berners-Lee behind it (used in https://solid.inrupt.com/)
• Things that need investigating:
  • Authentication/authorization
    • API can be accessed, but also through local services like a web app and third-party integrations
  • Combining multiple APIs
    • Is there one entrypoint or many?
• JSON-LD context resolution needs to be considered (at least HTTP caching, but ideally using local context files, like https://github.com/science-periodicals/ontology/blob/gh-pages/context.js)