Open data and FAIR as a guiding lens

The Open Knowledge Foundation's Open Definition v2.1 specifies that data is open if anyone is free to use, modify, and share it for any purpose. This deceptively simple definition breaks into three dimensions that must all hold simultaneously.

1. Legally open: released under a licence that grants permission to use, copy, redistribute, and adapt with no or minimal restrictions. Placing data on a website without a licence does not make it open: copyright applies by default, and no one has explicit permission to do anything with it. A dataset marked "for academic use only" is legally closed even if publicly accessible.
2. Technically open: published in a format that can be processed without proprietary software. A PDF containing a table is legally open under OGL but technically closed: the data cannot be extracted and analysed programmatically without error-prone parsing. CSV or JSON under the same licence is both legally and technically open.
3. Practically open: discoverable, well-described, and usable without significant investigative effort. Data published on an unlisted FTP server with no metadata, no data dictionary, and no contact information is legally and technically open but practically closed: most potential users will never find it, and those who do cannot interpret it without extensive reverse-engineering.

The FAIR principles were published by Wilkinson et al. in Scientific Data (2016) as a framework for maximising the reuse of research data. FAIR is not a licence framework or a file format requirement. It is a set of properties that data and its metadata should exhibit to support automated discovery and reuse by both humans and machines.

Findable: Data and metadata are assigned a globally unique, persistent identifier (DOI, ORCID for people, ROR for organisations). Metadata is indexed in a searchable resource. The identifier in the metadata resolves to the data. This addresses the primary reason research data goes unused: it cannot be found.

Accessible: The identifier resolves via a standardised, open, free, and universally implementable protocol (HTTP, HTTPS). Metadata remains accessible even if the data itself is embargoed or restricted. Accessibility does not mean the data must be free or publicly available: it means the access mechanism is open and standardised.

Interoperable: Data uses a formal, accessible, shared, and broadly applicable language for knowledge representation (controlled vocabularies, ontologies). Data includes qualified references to other data. This enables automated integration across datasets from different sources without manual mapping.

Reusable: Metadata richly describes the data with a plurality of accurate and relevant attributes. Data is released with a clear and accessible data usage licence. Provenance is documented. Data meets community standards for the domain. This is the terminal goal: data that can be reproduced and combined with confidence.

Tim Berners-Lee proposed the 5-star deployment scheme in 2010 to provide a progressive framework for publishers improving their open data quality. Each star level adds to all previous levels.

• 1 star: Data is available on the web in any format under an open licence. A scanned PDF of a table under OGL qualifies.
• 2 stars: Data is available in a machine-readable structured format (Excel spreadsheet rather than scanned image).
• 3 stars: Data is in a non-proprietary format (CSV, JSON, XML rather than XLS). No specialist software is required to open it.
• 4 stars: Data uses W3C standards (RDF, SPARQL) so that others can point to it using URIs. Each data element has its own stable URL.
• 5 stars: Data is linked to other people's data to provide context (linked open data). The dataset cites and connects to related datasets using shared identifiers, enabling automated graph traversal across data sources.

Most government open data in the UK operates at 3-star level (CSV under OGL). The UK National Biodiversity Network Atlas and some components of data.gov.uk achieve 4-star level. True 5-star linked open data remains rare outside academic and semantic web contexts.

Licence selection determines what downstream users can do with data. Mismatched licences block data combination: a dataset under CC BY-SA 4.0 cannot be legally combined and published with a dataset under CC BY-ND 4.0 because the output cannot simultaneously satisfy both licence conditions.

Open Government Licence v3.0 (OGL v3.0) is the UK public sector standard for Crown copyright data. It permits use, adaptation, and redistribution for any purpose provided attribution is given and the OGL notice is preserved. OGL is compatible with CC BY 4.0 and CC BY-SA 4.0. Ordnance Survey OpenData, Companies House data, and most ONS statistics are released under OGL.

Creative Commons Attribution 4.0 (CC BY 4.0) permits use, adaptation, and redistribution for any purpose including commercial, provided attribution is given. It is the dominant licence for open research data and is compatible with OGL v3.0. It imposes no share-alike requirement, meaning derived works may use different licences.

Creative Commons Attribution-ShareAlike 4.0 (CC BY-SA 4.0)adds a copyleft condition: derivative works must be published under the same or a compatible licence. This prevents proprietary lock-in of derived datasets but restricts combination with non-share-alike materials.

Creative Commons Zero (CC0) is a public domain dedication. The publisher waives all copyright and database rights to the extent permitted by law. No attribution is legally required, although community norms still expect citation. Wikidata, most US government federal data, and many genomic reference datasets use CC0. It is the least restrictive option and creates the fewest downstream combination problems.