#9 Configurable Metadata

Institutions

Cornell - Lynette

* Project to create collections; Each collection should have additional metadata fields for that project
* Different fields by type
* Fedora 4

University of Alberta - Matthew

* Very specific metadata for their technical reports; Theses need different data
* Fedora 4

Indiana University - Daniel

* Implemented flexible metadata; Randall gave a talk at HydraConnect 2015 (link please)
* Most on Fedora 3

University of Notre Dame - Jeremy

* Thinking of platonic form of metadata; Yet each object is its own thing
* Fedora 3

Oregon State - Ryan

* Similar needs to the above; Not all fields are applicable to collections
* Editing templates per collection
* Fedora 3

Stanford - Joe

* Lots of legacy data; In deep with lots of data
* Multiple inputs - Administrative interface, self-deposit interface
* Perhaps copied code
* Fedora 3

UC San Diego - David

* We have monstrous amounts of metadata that can be applied to all objects
* Certain objects can have a subset of those fields
* Custom repository

Concerns

  • Concern with renaming metadata field under a new name
    • How do we make data migrations
    • Has metadata declaration if you allow a form driven interface
  • By Collection (User Collection / Admin Set)
  • Concerns for flexible metadata
    • Indexing
    • Discovery
    • Browse
    • Create/Update
    • Faceting
  • Discussion about longitudinal stewardship of data
  • If we are going to configure this how would we go about doing this
  • Integration with class definitions that inherit from ActiveFedora
  • External configuration file; A schema.rb if you will; Or many configurations
  • Persist configuration in Fedora; the master copy is Fedora
  • Persist the configuration in a database
  • Configurable at runtime vs. boot time
  • Is the "metadata schema" an application or is part of the data and object
    • All of the data may not be consistent with the schema at a given moment
  • We need schema enforcement of the configuration
  • Repeating class definitions (for slight variations) is indicative of a problem; We shouldn't be doing that.
  • Discussion concerning I18n and how to represent this information
  • Configuration for terms/predicates should include I18n keys

Composition

By focusing on modeling predicates (the atoms of our objects), we can compose molecules. If a new predicate comes along, if we use a common configuration (and interface) we can share rendering strategies for predicates.

How do we combine configurations; Simple union as per order of application's configuration lookup.

Design Notes