DSpace Architectural Review

Notes from Thursday, 26 Oct 2006 (JSE)

I. Review of Agenda

1. Data Model Revisit (aggregation, etc)
2. Concrete data model and storage
3. Larry Stone: History, Provenance, Audit
4. Etc...

II. Abstract Data Model Revisit (Rob)

• summary: fixing the "bundle" system
  • See RJ notes

See diagrams of Rob's draft proposals

• version 1
• version 2 (typed)
• version 3 (typed)

Richard Rogers proposal

• "strawman 2"

(MS) Observation: Rob's makes more explicit how maps to FRBR

• whereas RR's more RDF- (or Fedora)-like

Need to test proposals against concrete examples

• See wiki, Data Model Use Cases
• See wiki, Asset Store Use Cases (older)

Problem with "Museum"-like repositories

• e.g. in which there are many representations of a "work"
• like many photos of a physical object
• (ms) FRBR vs VRA

(jse) Separate data abstraction from application

• (md) jsr-170 node/property model
• (rj) let's be clear on what we need to worry about; this talk og nodes/properties is implementation...

RECOMMENDATION: (draft) Adopt Rob's Item/Representation/File (with Metadata for each)

• ...and publish application note/"best practices" on meanings
• change: Bundle to Representation...
• change: Bitstream to File...

Example: "submitting a pdf with some metadata and a license"
*item is an article + descriptive metadata
*manifestation is a pdf
*file is a pdf

problem in linking-in files as metadata

LUNCH

III. History, Provenance, Audit

Overview of Larry Stone's stuff by RR

• See Larry Stone's Event System prototype
• ...and DSpace Event Mechanism

1. Event Mechanism

2. History needs to be based on logical events (transactions) not DB events

3.