Member Node Prototype API
-------------------------

- get(token, ID)
- synchronize(token, ID)
  - trust required - need to verify identity of caller
  - should only be called by CN
  - for prototype, the trust can be specified at the network level (buy time for feedback from distributed identity group)

- heartbeat()
- listObjects(date...)
- getSystemMetadata()
  - returns a subset of the system metadata


Coordinating Node Prototype APIs
--------------------------------


- synchronizeCompleted() 
  - taskStatusUpdate()
  - completed, failed with error, still in progress, pct complete, timed out

- reserveIdentifier(ID) -> T/F
  - reserves identifier for pending retrieval of new object from MN

- getSystemMetadata(token)
- search(token, query) -> resultset

- resolve(ID) -> location(s)

- login()
  - username, password or certificate
  - DAAC data retrieval requires username passwd.  trusted location for submission of metadata
  - define query syntax
  - resultset = list of IDs
  - three roles: admin, anonymous, authenticated
  - token should include some encoding of the user identity
    - potential privacy concerns
    - explicit privacy agreement  -> governance group

Versions
--------

Two streams of development.
a. implements:
   - shared search
   - replication of data and metadata
   - 3 CNs synchronized
   - no authentication or access control
   - metadata interoperability (harvest and make accessible through search)

b. implements:
   - a) plus
   - authentication
   - access control
   - 


V0.3 (minimal internal prototype) 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Aim for February 16 minimal functionality

Functionality

  - CNs read content from MNs
  - CNs implement resolve()
  - CNs synchronize content (science metadata) between themselves
  - System Metadata defined
  - Search using Mercury
  - client library (proto ITK)
  - logging (debug style)

Tasks
  - general documentation cleanup
  - clarify API documentation
    - for listObjects(), evaluate correspondence with OAI-PMH

    - comms = REST(ish), but RPC (e.g. protobuf) can be simpler
    - define REST URI encoding patterns for all operations (e.g. Serhan Akin)
    - Define error messages and handling for all APIs

  - Review subset of use cases that apply here and ensure that all API calls are considered
  - Review list of use cases for applicability (1, 2, 6, 9, 24, 10, 16) and map to Feb and July targets

  - set up unit tests for APIs

    - testing MN implementation of APIs
    - Testing CN implementation

  - build stubbed out MN APIs

    - generate client side and server side stub libraries for CN and MN
    - java, perl, python

  - MNs implement get(), heartbeat(), listObjects(), getSystemMetadata()

    - ORNL DAAC
    - Dryad
    - KNB
    - CDL
    [-NBII]

  - Spec hardware requirements for CNs

    - order + deploy hardware where necessary
    - evaluate deployment on EC2 (Eucalyptus?) Budget constraints?  

  - Deploy metacat + mercury on VM images

    - Adapt mercury to use metacat documents
    - Trigger mercury to update index
    - Dependencies
    - Incompatibilities?
    - Implement CN API methods
    - Architecture details of hybrid
      - message bus, dispatching calls to metacat | mercury
      - Mule-ESB

  - Network level security (CN-CN and CN-MN)
    - IP + host based?
    - Certificates?

  - Evaluation of search possibilities (e.g. THUMP, SRU, OpenSearch, SPARQL, XQuery)
    - What are the requirements for search systems over time
      - kw over metadata -> sophisticated search over data
      - What existing query languages map well to the search requirements
    - Generate a draft document that covers the range of available query languages / systems
      and their utility towards providing suitable search + discovery for scientific use of DataONE

  - Trajectory / longer term goals for search operations (e.g. search of data)
    

V0.5 July 31 2010
~~~~~~~~~~~~~~~~~

- MNs implement synchronize()
- CNs implement taskStatusUpdate()
- CN scheduler for directing MN replication operations
- System metadata replicated between CNs
- API for search standardized (cross mercury and metacat)
- logging


V0.9
~~~~

  - Authentication
  - Access control
  - Re-engineer problems in previous iterations
  - more MN types
  - ITK 
  - event notification / subscription


V1.0 (Year 3 Public release)


V2.0 (Final year 5 deliverable)