Linking research & learning technologies through standards

The Global Registries Initiative (GRI) is a collaboration between the Australian National Data Service (ANDS) and counterparts in the UK (IESR) and USA (OCKHAM Project). The initiative aims to make it easier to discover research data and publications stored in disciplinary and institutional repositories. It will do so by centrally registering national registries of research data, and the service points for interacting with those registries. Clients can then be configured automatically to search across all registered service points: the registries are formed into an ad hoc federation, which can be used by aggregators, mash-ups and portals, to improve access to scientific resources wherever they are stored.

GRI’s goals are consistent with those of ANDS, which aims to improve access to Australian research by gathering descriptions of research data collections from projects and institutions throughout the country, and making them available for search through a central portal. GRI extends the scope of discovery, by federating registries that do what ANDS is doing in Australia—just as a repository federation makes the content of all its repositories discoverable, without the content of the repositories necessarily being stored in the one central spot.

The GRI partners met in November in Canberra for a two-day workshop, to plan its work programme for the next couple of years. The meeting outcomes are given on the GRI website (PDF); the following is a summary of what was discussed.

• Registries of data can be modelled hierarchically: datasets are registered locally, but are most effectively searched for globally. To enable global discovery, data descriptions are aggregated up to the national level, or even the global level in discipline-specific registries. But no single registration body can register all the world’s scientific collections. Instead, GRI envisions a federation of national and disciplinary registries, with an interoperable core of services based around a registry-of-registries. GRI itself is merely a directory of registries, with minimal core functionality (including WSDLs and machine readable descriptions): it does not store information on individual collections or resources.
• GRI’s use lies in exposing that directory to third party discovery tools (portals, applications, catalogues, mashups), of which LibraryFind is an exemplar. The discovery tools, which GRI will register and link to, can then perform their own harvests or federated searches on the registries they discover via GRI. Developing discovery portals is out of scope of GRI, and is open to all GRI users. GRI will encourage national project funding agencies to promote researchers registering their datasets with a global registry either directly, or via a national or discipline registry.
• GRI’s scope includes both scholarly information (publications) and data sets. GRI’s distinctive value compared to similar projects like OAISter are its authority, and its support of descriptions for both collections and service points in registries. GRI will enable third parties to tag collections to make them more discoverable. However, no one universal tagset can be imposed across all scholarly domains, and GRI will not attempt it.
• GRI will formulate a shared descriptive model of registries, including the network location of collections, and their service offerings. Access rules will remain the domain of target collections, realised through their service layer; it will not be captured by GRI. GRI will not formulate its own standards, but will pursue community profiles of established standards. It is critical to establish more concrete use cases for how the GRI registry will be used, from the various potential user communities; this will proceed independently.
• There is a risk of granularity confusion in what level of collection the GRI registers: the registered collections may themselves be registries of registries. GRI cannot prescribe what it registers, since that must be driven by the community; but it will record and publish to users the different granularities of registries that it has registered.
• GRI will seek endorsement from a range of international bodies according to their distinct strengths. GRI will seek to expand its membership, including bodies which already publish registries-of-registries with service offerings.
• GRI will support the OAI-PMH standard for harvesting and the OpenSearch standard for (federated) search. It will support the full OAI-PMH protocol, including deleted records and incremental harvest. Participating registries must support OAI-PMH—though not necessarily the full protocol; and may also support OpenSearch.
• The data model that GRI will use for registry description will be derived from the IESR data model, and will be kept very simple since it needs to apply across a broad range of registries. Most of the work on the data model will involve profiling rather than modifying the existing schema. The founding GRI partners all use variants of ISO 2146 for registry description, so they are broadly consistent.
• GRI will establish a permanent secretariat for the next 18 months, based at ANDS, and will review sustainability options after that point.