From W3C Wiki
Jump to: navigation, search

White paper on "Vocabulary Management"


This white paper was initiated during breakout discussions at the 2009 Dagstuhl seminar "Semantic Web Reflections and Future Directions". Participants in this breakout session were Chris Bizer, Dan Brickley, Stefan Decker, Aldo Gangemi, Asun Gomez-Perez, Harry Halpin, Tom Heath, Ivan Herman, Natasha Noy and Guus Schreiber. Notes from this session are at [1].

URI Scheme

Versioning strategy

There is currently no "best practice" for the preferred versioning policy of vocabularies. For the moment, vocabulary owners should probably prefer one of the following current practices:

W3C document versioning policy

A W3C document such as the SKOS Recommendation has two types of URIs:

  1. One basic URI, which always points to the latest version. For the SKOS Recommendation this is [2]. This URI contains no version information such as dates or numbers.
  2. One or more version-specific URIs, which typically include the publication date of the version as an identifying part. For example, the URI [3] is the version specific URI of the official SKOS Recommendation, whereas the URI [4] is the version-specific URI of the SKOS Proposed Recommendation document.

Version URIs are connected to each other through a `previous-version` link; this link always defines which version directly precedes the version at hand. The result is a chain of backward links to the first version published (i.e. the one with no `previous-version` link). For the SKOS Recommendation the first version is [5].

The `this-version` link is used to connect the basic URI to the corresponding version URI.

 Note: Although the W3C scheme is originally meant for documents, it is also well-suited for vocabularies. Version numbers could be used in version-specific URIs as an alternative for dates. 

Dublic Core versioning

DCMI (Dublin Core Metadata Initiative) has a more extensive versioning or "History Model" for their own DCMI term set. The main difference with the W3C scheme is that the DCMI scheme allows for recording meta-information about the versions. The main features are:

  • A version can have a `replaces` and/or a `is-replaced-by` link with another version.
  • A version can have a `status` property with values such as `conforming`, `recommended` and `obsolete`.
  • A version can be linked to a `Decision`, which describes the rationale for the new version.
   Note: The DCMI history model is applied to DCMI terms, i.e. vocabulary elements. However, there seem to be no obstacles for using it for a vocabulary as a whole. 

The URI scheme for identifying the version of DCMI terms is similar to the W3C scheme. For example, the URI [6] is the basic URI for the DCMI term `hasPart`. The `dcterms:hasVersion` property link identifies the URI [7] as the most up-to-date version-specific URI.



Minimal attribution metadata


DNS Redirect advices, examples, templates

@@Here should go stuff not covered in the Recipes note

Sustainability, physical hosting, backup policies

Ideally, vocabulary services should be hosted by organizations of which the management has expressed commitment to long term support of these services, keeping the URIs of the vocabulary terms persistent. Note that "persistence is a function of organizations, not technology"