Permalink URI: https://w3id.org/id-rules/1

The final rule describes referencing recommendations for data redistributors: data aggregators, who collect information from different sources and re-display it; data publishers, who disseminate scientific knowledge through publications; and online reference material.

When database identifiers are referenced in narrative online text, they should always be hyperlinked to their URIs or to metadata containing their URIs [e.g. 19, 20]. Where cross-references are displayed to humans (e.g. as in Monarch disease overview pages) consider using a CURIE notation whose prefix (see Rule 2) is as assigned by the data provider. Where machine parsing is intended (e.g. Monarch metadata landing pages), any CURIEs must be given together with a machine-parsable definition of the prefix-to-source mapping e.g. Monarch CURIE map. Access URLs are volatile (see Rule 4) and must not be used for referencing. This is especially relevant for bundled and transitive references to identifiers, such as those found in published bioinformatics pipelines (for example, Research Object or Galaxy).

Where there exists no provider-issued or commonly-used prefix for a dataset, the data distributor should select one, preferably after discussion with the data provider. Similar documentation best practices apply to data providers and redistributors (Table S6).

Redistributors of data should monitor their references to other sources; any ‘dead’ links should be reported to the original data provider. If the original provider does not fix the broken link, the reference to it should be marked obsolete both visibly (for user interaction/interpretation), and within any accompanying metadata (for computational interaction/propagation).