Executive Summary of Results from Report by Hoebelheinrich and Janee

From Earth Science Information Partners (ESIP)

From Report to ESIP Federation Data Stewardship Committee on the Implementation of Selected Identifier Schemes to Earth Science Data Objects as part of its Identifier Testbed Activities - draft of 6 January 2014

In hands-on testing the ESIP Federation’s Data Stewardship Committee Identifier Testbed Project found that Digital Object Identifiers (DOI), Archival Resource Keys (ARK), and Handles are the most suitable identifier schemes to serve as unique locators for Earth system science data and information. Implementation of these three schemes demonstrated their:

  • ease of use, both for human users in creating and managing individual identifiers and for programmatic clients in operating on batches of identifiers via APIs;
  • support, in terms of documentation and infrastructure, and of the stability of the backing organizations;
  • widespread adoption by different communities;
  • scalability with respect to both the number of identifiers that may be created and the rapidity with which they may be created.

DOI and ARK identifiers additionally offer support for citations, and publishers have expressed interest in harvesting these identifiers and their citation metadata into search systems. Handles require more investment, particularly in that the client organization must run a specialized, dedicated, local server. By contrast, DOI and ARK identifiers are accessed via centralized, global systems.

The ID Testbed Project found that Universally Unique Identifiers (UUID) are most appropriately used as unique identifiers, and are the most appropriate scheme for that purpose. Life Science Identifiers (LSID) are possibly suitable as unique identifiers, but they are unsuitable as unique locators due to their incorporation of domain names. Additionally, LSID suffers from relatively low adoption by a narrow community.

Persistent Uniform Resource Locators (PURL) offer no means for creating opaque identifiers, and the API support for batch operations is poor. Object Identifiers (OID) and Extensible Resource Identifiers (XRI) are least suitable for Earth system science data. The XRI scheme has yet to become fully operational. OID identifiers are really targeted at other information types (controlled vocabulary terms, for example).