FAIR Publishing — PURLs, Content Negotiation, LOV

FAIR Publishing — PURLs, Content Negotiation, LOV

Stable IRIs, HTTP content negotiation, registration in LOV / BioPortal — make your ontology actually reusable.

Why it matters

An unpublished ontology is a private notebook. FAIR is the difference between 'we built one' and 'three other teams adopted it'. PURLs + content negotiation + a public catalogue listing are the minimum.

Going deeper

The FAIR publishing checklist for an ontology release — do all five or you're not really FAIR yet:

1. Persistent, owned IRI — use w3id.org or PURL so the IRI doesn't change when you switch hosts. Direct GitHub Pages URLs are not persistent.
2. Content negotiation — the IRI must serve text/turtle, application/rdf+xml, application/ld+json and text/html (a human-readable doc) from the same URL. Widoco generates the HTML + Apache rewrite rules for you.
3. Versioning + changelog — every release exposes its owl:versionIRI and a public changelog (see the Versioning lesson).
4. Registration — list in at least one catalogue: LOV for general-web vocabularies, BioPortal for life-sciences, OBO Foundry for coordinated biology, AgroPortal for agriculture.
5. License — an explicit dct:license triple (CC-BY 4.0 is the safe default). No license = no legal reuse = not FAIR.

Most 'we built an ontology' projects ship 1–2 of these and call it published. Shipping all five is what separates the projects that get adopted from the projects that get forgotten.

FAIR publishing is giving your ontology a postal address and a building directory — instead of leaving it in a desk drawer.

A house in a desk drawer is real, but unreachable: no postcode, no street name, no listing. A house with a stable address (PURL), a doorbell that answers differently to the postman, a tradesperson and a tourist (HTTP content negotiation: Turtle for the SPARQL engine, RDF/XML for the legacy tool, HTML for the human visitor), and a listing in the city directory (LOV, BioPortal) is one that anyone in the world can find, ring, and use.

Almost every dead ontology you'll see in academia died from the desk-drawer problem — the model was good, but the address kept changing, the doorbell only answered in one format, and nobody listed it in the directory. The FAIR principles are nothing more exotic than the postal service of the Web.

Tools & resources

• GO FAIR principles — the canonical reference: https://www.go-fair.org/fair-principles/
• w3id.org — community-maintained persistent identifiers (free, GitHub-managed): https://w3id.org/
• PURL.org — the original persistent URL service: https://purl.org/
• Widoco — generates HTML documentation + content-negotiation config from an OWL file: https://github.com/dgarijo/Widoco
• Linked Open Vocabularies (LOV) — the catalogue for general-web vocabularies: https://lov.linkeddata.es/
• BioPortal — life-sciences catalogue: https://bioportal.bioontology.org/
• OBO Foundry — coordinated biology ontology catalogue + acceptance criteria: https://obofoundry.org/
• W3C TAG — Cool URIs for the Semantic Web: https://www.w3.org/TR/cooluris/