Fedora: A service-oriented repository architecture to meet the challenges of the "Info Grid"
Sandy Payette
Co-Director Fedora Project
Cornell University
Presentation: ppt (1.8M); pdf (2.5M)
Traditional roles for digital repositories
- collections management, scholarly publication, document discovery, storage and archiving
- these roles are evolutionary, not revolutionary
- document-centric
The future: Rich Information Networks
- documents
- data
- services
- actors
"This is the information grid, fully realized"
Signs of Change
- grid computing in sciences
- humanities computing: hyperlinked historical documents, online sites, tibetan digital library
Need to translate the broad info grid requirements into repository architecture.
- rich objects, dynamic objects, info integration, service-oriented repository architecture
- security and fine-granualarity access controls
Fedora
- XML, Web Services (SOAP/WSDL) as well as REST
- OAI, RDF, OpenURL
- version 2.1 coming
who
- NDSL
- OhioLink
- many others
Flexible Digital Object Model
Document = Aggregation + Annotation + Service + Data
use "disseminators" to connect objects to services
Benefits of Object Relationships
- Enable an "information network overlay"
[diagram of web services]
example
image/gif
disseminator: methods of images service (via WSDL)
disseminator: methods of metadata service (via WSDL)
Want Fedora to exist in a broader service framework.
[diagram of Fedora Service Framework (2005-2007)]
planning BPEL-based workflow engine (Fedora Workflow)
cross-repository
* Pathways Project
http://www.infosci.cornell.edu/pathways
- Next-generation scholarly communication system
* Pathways InterDisseminator Service
Fedora Security Architecture
- tomcat, ldap (pluggable modules)
- initial work on Shibboleth (developed by OhioLink)
- XACML-based policy enforcement: provides very flexible and granular control
- including controls on backend services
Fedora Devel Plans
- new services
- federated repositories and name service
- scalabilitya and performance
- tighter Shibboleth integration
- event notification - pub/sub
- new client apps
= FIRE client
== advanced scholarly workbench with workflow
= tools for RDF browse
Fedora and Data Grid
- Storage Resource Broker (SRB)
- SRB as alternate backend for Fedora storage
Ongoing Research Challenges
- enable low barrier to entry
- better integration of objects and services
= semantic service descriptions (OWL-S)
= object to service dynamic matchmaking
~ "we need to build easy tools, on top of the services"
- distributed security, identity and polcy
= interop of authorization policies
- preservation
= how do you preserve a distributed aggregation? a dynamic resource?
www.fedora.info
Q: How much time to bring up a repository?
A: technology is easy (hours), the rest depends on you
Q: when will clients for ingest be avail?
A: some now, but FIRE (more sophisticated) next year
Q: Sustainability - what happens when funding runs out
A: developing a consortium, looking at possible models
UVirginia will maintain the code base for 7 years from the end of the funding.
Followup: he believes you need vendors for long-term sustainability
mentions VTLS
Q: do you index content of datastreams?
A: No. That's really a separate service, use some other tool to do that.
Recent Comments