This list of issues results from discussions at the Distributed Services Registry Workshop, University of Warwick,14-15 July 2005. Thanks to all participants.
What is meant with distribution in the context of registries?
- General benefits and disadvantages of distributed registries
- What are the characteristics of existing approaches and example registries?
- Different degrees of distribution: From more to less cooperation:
centralized (central access)
federated (distributed creation, central access)
replicated (distributed access)
distributed interoperable (distributed creation and access, referrals. Cross-search)
distributed non-interoperable (distributed creation and access)
- Different types of registries: services, collections etc.; transactional vs.content services
- Different coverage: Global, institutional, subject registries etc.
- What types and coverage should one DSR entity comprise?
- How should the different types of registries interact?
Existing approaches of service registries: UDDI; OCKHAM
Example registries: Astronomy US Virtual Observatory initiative (Registries: Carnivore, NVO Reg.); OCKHAM; IESR; myGRID registry + OMII Grimoires registry project; OCLC?; DLF Aquifer project/UIUC. library vendors service registries (e.g. ExLibris)
Plans: TEL; UK GRID/E-Science Centre? Network of service registries; JISC E-Learning Framework and Virtual Research Environments; SCONE, Scottish collection description service (45 Z services, soon OAI services); Open DOAR
Distributed service registry: Framework and role
- What is the framework the registry operates in; is there a need for an OAIS like model?
- What is the role of registries in a Digital Library workflow and context, in the wider architecture?
Scope and approach
- What is the purpose of a DSR (evidence of need). What problem are we solving? Is it necessary or will Google fulfil all needs?
- Is a global DSR a realistic goal to aim for?
- What are the minimum requirements in different communities?
- What is the core functionality and what is added value (in applications)?
- Is the DNS a good and sufficient operational and functional model for a DSR?
- Usage: Who are the users and how will they (want to) use it? What are relevant use cases in different communities?
- Scoping: What is a service (e.g. a web site, a schema) and what should be registered?
- Service driven vs. collection driven approach
- Should service and collection description registries be separated?
- Current best practice vs. vision
Organisational and cost-benefit issues
- Who will fund a DSR? What is a proper cost model? Who owns the DSR?
- DSR as commodity or public infrastructure?
- Quality assurance, importance for trust and take-up
- What are the barriers to adoption?
- Who operates, manages and maintains DSR services?
- Who creates and updates the content?
- How to motivate contributors/content providers, initially and for updating?
- What is the value of a DSR? Who receives the value?
- Cost benefit analysis
Identifiers
- What is identified?
- How can the identifiers become unique and persistent?
- Identification vs. Location of services (for access)
Metadata
- Description purposes: Discovery, access control, IPR
- Different communities require different approaches
- How to deal with heterogeneity
- Collection Description: core with extensions
- Service description: what is needed for human and m2m interfaces?
- Annotation by third parties?
- Supporting multiple views
Appropriate record/service
- Who decides which record the user sees if there are several metadata records for the same service (appropriate record) or several services for the same purpose (appropriate service): the DSR or the applications? (?issue to be solved by DSR "cloud")
- Will de-duplication be desirable?
- Will part of the resolution be invisible to the end-user? (cf. DNS: real sites hidden under same domain URL)
Architecture, Protocols
- What is the relationship to service oriented architectures?
- How to interface to other shared services; how to choreograph multiple services?
- Functional requirements for a DSR architecture; what is the core, what are external applications?
- What is the most suitable topology (each node might be a "cloud"): hierarchy or P2P? How to discover DSR nodes?
- Replication (of content) vs referral (like DNS) vs broadcast search
- What are suitable protocols for: notification (of new nodes, new content);
selective sharing (incl. Harvesting)?
- How does one deal with scalability problems (e.g. registry cross searching in the OCKHAM approach)?
- What provisions are needed for searching and caching?
- ? What are the most suitable standards to store (and access) the DSR service data? (UDDI, Explain Lite, Zee-Rex)
Access control/IPR/licensing for DSR records
- IPR wrt metadata records and aggregation, commercial use; licenses (Creative Commons?)
- Different access modes to the DSR
- Right to enter the DSR "cloud"
- Authorization/authentication needed for different parts of the metadata, public and private metadata; different access rights to different sets/parts of the metadata
- How are the rights expressed in m2m form?
Other issues
- What is the potential role of ebXML?
Ways forward
- Ways forward for the service registry community?
- Can we expect sufficient cooperation? Can communities buy in to a common metadata standard and DSR architecture?
- Implications for the IESR
