FedoraShare Session
Missing - Call disseminations using REST API
Web 2.0 - Add your content easily.
Relate the stuff from multiple sites.
What is the user experience?
Eddie - Goal is Fedora API improvements or FedoraShare application?
Inform, Include, Provide easy simple application (Sandy - note Timeline from Simile)
YouTube as a dissemination. (example dissemination)
Google has a different OpenID impl.
With FedoraShare HTML page with YouTube embed tag.
Yahoo Pipes? Investigate. Kepler.
Simple back-end.
Users as first class objects.
Messaging does API-A.
Dan - Rebuild slider in FedoraShare.
Triple benefits.
Replication into multiple sources.
Flex is well supported.
Low impact frontend.
Dan - Local App/Server
Paul - Apple Cappucino.org written in Javascript as alternative
Developers - My shop does not support?
Reusable code?
Separable components.
Mulgara
- JTA (Andre)
- Handle all transactions
- One write at a time - need to move to multiple writers
- Connection API
- Graph URIs (permits arbitrary URIs - does not identify host)
- Rmi://host.com/server1?graph=foo:bar as opposed to # (fragment ID)
- SPARQL Engine
- SPARQL Language
- Focused on join optimizations rather than filtering which is needed for complex SPARQL queries
- Meeting W3C SPARQL spec especially what goes in must come back exactly (cannot be canonicalized)
- Sesame Integration
- James at Sesame giving good feedback
- Sesame is 100% SPARQL but not Mulgara. SPARQL has distinct clause which causes problems with set semantics. Not easily fixed. No support for Regular Expression (different from Java). Order-by arbitrary expression support.
- HTTP APIs
- TQL (iTQL name going away)
- SPARQL
- Using Servlets
- Throw exception if "insert" into Get operation
- Post expects an RDF file
- May be output in JSON
- HTTP UI
- Cleaned up no longer implemented in Barracuda
- Tutorial mentions graph URI: (not needed for real) Select Query: Query: Submit Button
- XA 1.1
- Scalability
- Multiple writers
- No problem with recycling Identifiers on 64bit systems
- 192 to 32 files for storage
- Better block writes
- Get test info from Ronald/Paul for presentations
- Change term from Model to Graph (W3C)
- Can index on all three orders SPO, POS, OSP
- Can index SPOG,POSG,OSPG,GSPO,GPOS,GOSP mostly these
- RLog - RDF Prolog
- hasUncle(X, Z) :- hasFather(X, Y), hasBrother(Y, Z). head :- body
Wiki Markup (X Y Z) y(X, Z) \[foo:bar rdf:type foot:person\]
- Foof:Person(foo:bar).
- Supports transitivity, symmetry, will support consistency
- User transitivity to implement a resolver
- JLine - Library for Java but must use native code to do it. Supports a lot of platforms. Andy wrote
Admin Client
- Minimal Version
- Public Comment
- Alpha Distribution
- Keep the old version for at least one iteration
funAPI
- Make Fedora Zotero friendly
- Easiest to use OAI-PMH
AtomPub
- Potential input servlet
- Simple get/post to URL
- How do we want to set all the other properties on the object
- Separate from SWORD, they may do the Fedora implementation themselves
Zotero as Client of Fedora
- Uses WebDav, Raymond, does he think they will support Atom PP.
- Do we want to rethink WebDav
Matt
- Look at same platforms, MS Silverlight too small installed base, using Flex and ROR and simple HTML.
- Cannot get full management info on datastreams
- Timestamp checked for lazy lock
- Cannot do relationships
- No getDatastreams
- Setting the content model
- Indexing (e.g Search Engine)
- Query by Relation (RISearch)
- Find by Content Model hard to do
- Find by Relationship
- Directly support the DC datasets
- Find by DC Terms
- Find by a Specific Property or Value
- Leverage off OTM finding objects query language
Aaron
- In NSDL Wiki
- Move to FR30 to reduce pain points.
- OAIprovider does not have an understanding of NSDL Information model
- Makes it difficult to find related items
- Views are inconsistent
- Resource > Aggregator -> Aggregator> Aggregator path can be traversed, queried, or be a service itself
- Eddie - Repository should know its own structure without depending on client
- Chris - Also like an inferencing problem
- Both Content Model and SDef represent kinds of "typeiness"
- For access this is easier to see but what about create/ingest/write
- Could we generate Client code?
- What REST style is most effective for Fedora?
Rick - Plone - CMS
- Plone - CMS in Python
- Collections are captured searches
- Links (except the built-in generic isLinkedTo) are user defined strings
- Content Modeler widely used.
- Active content display server
Paul - Mulgara Roadmap
- Rules engine has been around. Compile Horn Clauses to rules.
- Has uptake.
- Distributed queries but need improvement. Can improve performance especially network.
- Cloud optimizations, wants help
- Clustering, multi disk optimizations (RAID 1, configuration)
- Hadoop
- Loading RDF is still the bottleneck but clustering is the coming problem
- SPARQL update, modify missing (delete then insert), no transactions
- Virtual graphs through resolvers, Mulgara does not have filters
- Rules, too course-grained for right now
- Consistency Checking
- RLog - needs to be more like a scripting engine, move to move parser then impl as resolver
- RETE engine for dynamic updates (most expensive)
- XA 1.1
- Security