GUID-1 Workshop Report
The Taxonomic Databases Working Group (TDWG) and the Global Biodiversity Information Facility (GBIF) completed their first Workshop on Globally Unique Identifiers for Biodiversity Informatics (GUID-1) at the National Evolutionary Synthesis Center (NESCent), Durham, NC, USA on Feb 1 3, 2006.
for a complete set of materials presented during the workshop. Download the report in PDF format?
A GUID framework is foundational in facilitating systems interoperability in biodiversity informatics. It meets the need for a universally adopted system for assigning and recognizing identifiers in the domain.
A GUID framework will help to manage and cross-link the many different types of entities that are manipulated analytically in biodiversity informatics and will improve interoperability with other related life sciences domains, such as bioinformatics and ecology.
The workshop delegates consisted of a representative cross-section of domain experts from around the world (see GUID1Participants
The goals of the workshop were to:
- Discuss the requirements for globally unique identifiers for biodiversity informatics
- Select an optimal GUID technology (LSID, DOI, Handles or other)
- Begin to identify key parameters for implementing an effective system
- Investigate the use of a RDF-based metadata architecture for GUIDs
- Form working groups to address key identified issues before the GUID-2 workshop
- Life Science Identifiers (LSID) seem the most appropriate GUID strategy in biodiversity informatics.
- The use of LSIDs does not preclude the use of other technologies where appropriate.
- LSID authorities must use the Domain Name Service (DNS) to support identifier resolution. (The LSID specification allows for other resolution mechanisms, but DNS is currently the only mechanism in use.)
- Although it is not possible to prevent multiple data providers from issuing alternate identifiers resolving to the same data record, the community should develop processes and tools to coordinate issuing of single identifiers for some classes of data (e.g. taxon names).
- Metadata should be provided as RDF serialized as XML and should exploit existing vocabularies such as Dublin Core wherever these are in wide use.
- The LSID getData method should be used only where it is possible and appropriate to return an unchanging series of bytes. In other cases only the LSID getMetadata method should be used. (This reflects the use of the terms "data" and "metadata" in connection with LSIDs.)
The main criteria leading to the selection of LSID
- The cost-model of DOI. That technology is predicated on the idea that a revenue stream can be constructed for the identified objects, typically sufficient to defray the cost. That this is not the case for most, if not all, of the objects that are likely to be identified in our systems.
- The more dynamic nature of LSIDs, which does not require prior registration of every individual identifier before use.
- The open nature of the LSID protocol and software stack, and the ease of implementing LSIDs on different platforms.
The group compared the GUID technologies according to the following criteria:
: Is the identifier free from embedded semantic information?
Opacity was identified as a possibly important criterion in that genuinely opaque identifiers could not be used to make false inferences about the object represented by a GUID. Handles, DOIs
all include similar levels of embedded information.
: Is there a body that monitors the assignment of identifiers?
has a more formal governance model for identifiers than the other standards. Assignment of identifiers is a more strongly contractual matter and all identifier assignment and access is mediated through the DOI
registration infrastructure. Several use cases for GUIDs in biodiversity informatics require more dynamic assignment and resolution paths.
: Is there any guarantee that identifiers will remain resolvable other than the commitments made by the assigning authority (commitments which must be made regardless of which technology is adopted)?
The central DOI
infrastructure holds the registered identifiers and makes some commitments to host orphaned data.
Registration of assigning organisations
: Must institutions register before being permitted to issue identifiers?
Issuing authorities for Handles and DOIs
are registered centrally. LSID
resolvers must be registered in DNS but do not need to be identified to a central LSID
Registration of identifiers
: Must institutions register each identifier before use?
are only resolvable if they are known to the central authority.
: Do the identifiers have a standard association with metadata?
have mechanisms to provide access to metadata.
: Does the identifier include a mechanism to retrieve the associated metadata and data?
are all resolvable in this way.
: Is there a commitment that the identifier will uniquely identify a single object?
all involve commitments to global uniqueness.
: Can an organisation's identifiers be transferred for a different organisation to resolve (e.g. upon closure of the issuing institution)?
An assigner of Handles, DOIs
can pass responsibility for resolution to another resolver organisation.
: Can individual identifiers be transferred for a different organisation to resolve?
may be assigned to other organisations to resolve. This is not possible with LSIDs
: Does TDWG have the ability to take over ownership of the standard and software if others stop supporting it?
Handle and DOI
are both based on proprietary technologies. LSID
is based on a more open strategy.
: Is the technology affordable for TDWG, GBIF and its partners?
TDWG partners together expect to assign many millions of GUIDs and have no model to fund the cost of DOIs
. The cost of licensing Handle technology is unclear. LSIDs
will involve costs in development of processes and infrastructure, but TDWG has more control over the process.
Summary Technology Comparison
The following table includes catalogue numbers and taxon names for comparison as these are examples of identifiers currently in use for data integration in biodiversity informatics.
|| Catalogue numbers
|| Taxon names
|| LSID |
|Registration of assigning organisations||-||-||+||+||-|
|Registration of identifiers||+/-||+/-||-||+||-|
Table 1 - Summary Comparison of GUID Technologies
There are still many issues to address before our community can fully implement an identifier system based on LSIDs
. The workshop addressed a number of specific issues and developed working groups to address the following issues:
The Infrastructure Working Group
- Developing white papers to address best practices and key infrastructure questions.
- Prototyping activities.
This group was formed to address the key issues regarding the deployment of LSID
as the GUID technology for biodiversity informatics. The mandate of this working group is to identify required or desirable policies and infrastructure components to ensure robust, long-term operation of shared GUIDs.
The following activities were identified:
- Specify minimal standards (including tools and services) for GUID issuance.
- Investigate long-term archival of LSIDs and associated data and metadata.
- Investigate establishing (optional) central registration authority.
- Investigate establishing repository for data and (orphan) datasets with GUIDs.
- Investigate the feasibility, existing actors and requirements for a "Publication Bank", i.e, a resource to act as a central registry of taxonomic literature and its digital representations, including assigning GUIDs to each publication.
- Clarify the distinction between GUIDs assigned to data objects and to conceptual entities.
- Investigate 3rd-party annotation and link-out mechanisms.
- Develop materials to communicate with wider community.
- Develop best practices for assigning resolver namespaces for LSIDs.
- Perform review of LSID specification to identify possible enhancements.
- Perform gap analysis of LSID software.
(Click the links above to view more information about each activity, including current status.)
The outcomes of this group will be a series of white papers addressing the key infrastructure issues. Those will be reviewed during the second GUID meeting later this year.
The Prototyping Working Group
Our community must experiment with LSID
technology and Ontology Engineering if we are to implement a production quality LSID
system, The working group will develop prototypes of test cases to test aspects of a GUID infrastructure.
The group will develop test LSID
resolvers using data objects provided by each domain, such as names, specimens, and concepts. This activity will also help involve (and train) the community in developing appropriate RDF ontologies, leading to concrete recommendations and implementations.
The potential prototypes to be developed and respective Conveners are:
- LSID resolver for taxon names - developed by nomenclators using IPNI database and an RDF version of TCS-Names - Group responsible: Roger Hyam, Sally Hinchcliffe, Paul Kirk
- LSID resolver for specimens using DarwinCore? (also ABCD?) - Steve Perry.
- LSID resolver for taxon concepts by SEEK using TCS.
- LSID resolver for observations by SEEK using EML.
- LSID resolver for character data by Damian Barnier, Kevin Thiele
- LSID resolver for images: Greg Riccardi (Morphbank), Bob Morris
The taxon names resolver has the highest priority.
Prototypes will address one or more of the following (but may not be full implementations of an LSID
- Hardware and software (including LSID stack)
- RDFS/OWL vocabulary for domain
- Data mapping between local data store scheme and shared ontology
Other important tasks identified by this group are:
This group will have 3 months to work on the specified tasks before preparing for the second GUID workshop.
Next Workshop: GUID-2
The TDWG Infrastructure Project is planning a second GUID workshop in late May or early June, 2006. At the time of writing a venue has not been decided.
The second workshop should cover the following:
- Review of the material produced between the workshops by both working groups (prototypes and white papers).
- Summary of the lessons learned in the process.
- Identify open issues and devise specific work plans to address them.
- Draft concrete recommendations on GUIDs for production systems - in general and for each specific domain (names, specimens, concepts, images, etc).
Information about GUID-2 will be distributed as soon as possible.
How to provide your feedback on this report
Contrary to the nature of wikis (including this one), we would like this report to be an static document reflecting what was discussed and agreed during the GUID-1 workshop, rather than a continuously evolving document.
For that reason, we will lock this page for editing. If you have any questions and comments, please edit the GUID1ReportComments
Ongoing discussion should happen elsewhere in this wiki. We are currently working to reorganize the wiki to provide suitable places for those discussions.
Thanks for your understanding.