An Introduction to Data Modeling with Fedora

1
An Introduction toData Modeling with Fedora

• Thorny Staples
• Fedora Commons, Inc.

2
Fedora Abstractions

• Data objects
• Content models
• Behaviors of objects
• Policies about objects
• Relationships among objects

3
(No Transcript)
4
A data object is one unit of content
Persistent ID
DC
RELS-EXT
Reserved Datastreams
AUDIT
POLICY
1
2
Custom Datastreams (any type, any number)
n
5
Content Models

• Create classes of data objects
• Expressed as Cmodel objects
• A Cmodel object defines the number and types of
  data streams for objects of that class
• A Cmodel object binds to service objects to
  enable appropriate behaviors to be inherited by
  data objects

6
Optional Object Behaviors

• Data objects can have different views or
  transformations
• Sets of abstract behaviors that different kinds
  of objects can subscribe to
• Corresponding sets of services that specific
  objects can execute
• The business logic is hidden behind an
  abstraction

7
Service Definition Object
service subscription
Cmodel Object
service contract
data contract
Data Objects
Web Service
Service Mechanism Object
8
A behavior call is a URL that contains
Object PID SDef Name Method Name
For example http//lib.va.edu/fedora/get/xxx1/sd
ef1/getThumb Can also add parameter and a
date-time stamp to access earlier versions of
the object
9
Policies

• Machine enforceable expressions of rules, what
  they are applied to and who they affect.
• Who is affected can be defined in different
  authorization sources, such as LDAP services
• Rules can be as simple as allow or deny.
• Rules are applied to objects as a whole, any
  datastream, or a dissemination, as well as each
  API call and more.

10
(No Transcript)
11
Historical Census Object

• 1870 Aggregate Census file of the US
• A character data representation of the dataset is
  the master
• It has a datastream that is used to access an SQL
  database that is only accessible through the
  object
• The SQL data can always be rebuilt from the
  character data
• It has a DDI codebook which has descriptive and
  structural metadata about the object

12
Relationships Among Objects

• Describes adjacency relationships among objects
• RDF data of the form
• PID typeOfRelationship relatedObjectPID
• Can used to assemble aggregations of objects

13
Content Modeling Styles

• Atomistic objects have a small number of
  datastreams that are each expressions of the
  whole, with relationships to other objects
• Many objects
• Much more flexible
• Compound objects usually include many
  datastreams, including information about the
  whole and its parts
• Forces a mixture semantics of whole and parts
• More difficult to take advantage of the Fedora
  abstractions cleanly

14
Book Objects

• XML file is the main datastream that represents
  the book as a whole
• Using the atomistic approach, a book with 400
  pages would be 401 objects
• Using the compound approach, 1 object with as
  many as 1201 datastreams for image files and the
  book file
• Example

15
Objects Representing Aggregations

• Creating parent objects for complex resources
• Representing explicit collections
• Representing implicit collections
• Creating digital surrogates for physical entities

16
Explicit Aggregations

• The parent aggregation object has explicit
  references to the PIDs of its children
• These references can be relationships listed in
  the objects Rels-Ext datastream
• Or they can be PIDs embedded in an XML datastream
  that gives a more descriptive context and can
  explicitly order them
• Example

17
Implicit Aggregations

• One object that represents the aggregation as a
  whole
• information about its meaning
• Rules for how to find the members
• Any number of objects can assert an isMemberOf
  relationship to the PID of the aggregation

18
Collections can be expressed as implicit
aggregations
isMemberOfCollection
isMemberOfCollection
Resource Index
isMemberOfCollection
Collection Object