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.
Eddie - Yahoo Pipes? Rick - 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.
Paul - 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.
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
(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
Bill - Admin Client
Minimal Version
Public Comment
Alpha Distribution
Keep the old version for at least one iteration
Eddie - funAPI
Make Fedora Zotero friendly
Easiest to use OAI-PMH
Eddie - 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
Eddie - Zotero as Client of Fedora
Uses WebDav, Raymond, does he think they will support Atom PP.
Do we want to rethink supporting WebDav?
Matt - REST API
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 - NSDL Implementation in Fedora 3.0
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
Development Community, Processes and Infrastructure
Chris - covered how we are provisioning projects
Need HA solution
Keep delivery on SourceForge
Paul Google Groups has some bad perceptions.
Backups of Wiki, Jira
Rsynch SVN.
Eddie - Mercurial
Rick Plone - Collective
Paul - Google Code
Dan - Developers Pages and Experience, Wiki, CMS
Dan - Developers Forums
Dan - Developers Forums
Eddie - Using Maven
Chris - Still want a source release.
Chris - Akubra
Eddie - Is Honeycomb dead. Chris - It represents XAM compliant storage.
Eddie - Why is "quiesce" function required
Sandy - Concerned with so little metadata. Chris - Many implementations with varying capabilities but other example may not be so concerned with disaster recovery.
Sandy - Consider "last-modified-date" as required read-only metadata
Cloud Storage
Bill - Summarized Amazon EC2/EBS/S3 experiments to date
Reviewed - Peter Murray article on OCLC vs Amazon S3 Storage
Reviewed Amazon SLA
Eddie - What about Fedora Clusters (or Fedora Federation)
Wrap Up Actions
Finish Challenges
FedoraShare, Admin GUI Technology
Test Maven to decide to do wholesale conversion
Due diligence on Confluence Plug-in license
Investigate Mercurial
Backup of Topaz site
Do we integrate Mulgara more closely? What are the boundaries?