Contribute to the DSpace Development Fund

The newly established DSpace Development Fund supports the development of new features prioritized by DSpace Governance. For a list of planned features see the fund wiki page.

Primary Objective

The primary objective of this proposal is that the DSpace metadata registry be "naturally" extended to support a richer and more expressive "Metadata Schema". Technical Objectives of the proposal are to provide the following features:

  1. Capability to Define Metadata "Profiles" that may be assigned to individual DSpace Objects
  2. Capability to Define "subPropertyOf" relationships in place of the legacy ns.element.qualifier approach for more expressive inheritance and mapping to OAI_DC
  3. Capability to have "immutable" DC, DCTERMS and other "well established" namespaces to treat as sources for "subPropertyOf" assignments.
  4. Capability to to Restrict and Validate Existing DSpace Object Metadata based on the assigned "Profile". 
  5. Capability to Apply these profiles similarly to any DSO; Communities, Collections, Items, Bundles and Bitstreams, even Groups and EPeople.

Expectations for Backward Compatibility

This proposal is based on the premis that changes to DSpace metadata characteristics must be backward comparable and retain the same functionality as previously existed to ease transition for all existing users of the platform.  So many different functional areas of DSpace are reliant on existing metadata functionality that it is criticial that any changes in functionality also have well defined and scripted updates across releases.  Thus another very critical feature of this proposal is that this new Schema model should support the above features without significant need to transform existing DSpace Item metadata nor the registry itself.  

Conceptual Definition of "Schema"

The DSpace MetadataSchema registry was designed based on an outdated concept of "Application Profiles" and "Qualified Dublin Core" that predated the current DCMI Abstract Model.  Due to this, there are number of significant shortcomings to the current implementation.

  1. Namespaces are not really "Schema"
  2. Schema may be validated, however, there are no actual rules in DSpace "MetadataSchema" or "MetadataField" data models.
  3. Qualification does not effectively meet needs for use of alternative namespaces nor support any ability for programatic mapping to DC for exposing metadata in other namespaces within in OAI_DC.
  4. The Schema and Fields defined are insufficient to support attributes and rules for validation of DSpace metadata fields in relation to Item Submission or other methods of Deposit.

The current "DSpace Schema" does not meet the requirements that a Schema is traditionally used for.  Schema are traditionally used to define a scaffolding or framework of rules which actual content can be validated against. While the current MetadataSchema/Field does restrict what can be assigned to any item in DSpace, it does not provide any support for validation of these assignments, nor allow us to further define the encoding of the metadata values nor if they are required or not.  At this time, much if of the validation, rules and encoding is poorly assigned instead, at the UI/Presentation level in the DSpace Submission input-forms.xml file and only enforced in the Describe Step of the Submission workflow.

This proposal seeks to extend the definition of the DSpace Metadata Schema to include support of these features previously found only in the Submission input-forms.xml. Formaizing a strategy for metadata validation in DSpace that is a new core feature.

Content Models For DSpace (Extending on MetadataSchema and MetadataField to provide "Metadata Profiles")

Rather than the current MetadataSchema applying to the namespace of the metadata fields that are allowed by the entire repository.  It is instead recommended that this table be repurposed and expanded to support creation of "Named Profiles" that can be easily assigned to DSpaceObjects as Metadata or or Content Models.  In this case, typing would initially be based on:

These above types will be expressed through the addition of properties to the MetadataSchemaRegistry and MetadataFieldRegistry tables to provide the facility to expand on and add additional Schema.  Some Hypothetical examples of such schema would be:

Community or Collection ProfilesDescription
Document Collection ProfileMetadata Fields Appropriate to Describe Details of a Generic Document Collection for the purposes of a Finding Aid
Journal Issue ProfileMetadata Fields Appropriate to Describe Details of a Journal or Journal Issue for the purposes of a Finding Aid
Image Gallery ProfileMetadata Fields Appropriate to Describe Details of an Image or Multimedia Collection, allowing UI Hooks that are beneficial for Multimedia (Slide Decks, Light tables, Viewers, etc)
Item ProfilesDescription
Scholarly ItemMetadata Fields Appropriate to Describe a Scholarly Research Article
Website ItemMetadata Fields Appropriate to Describe a number of individual Bitstream files that constitute a website.
Thesis ItemMetadata Fields Appropriate to Describe a Dissertation of Thesis Item based on conventional ETDMS terminology
Technical Report Item 
Journal Article 
Learning Object Item 
Bitstream ProfilesDescription
Streaming Video ProfileMetadata Fields Appropriate to Describe a moving picture or video
Image ProfileMetadata Fields Appropriate to Describe an individual image
DocumentMetadata Fields Appropriate to Describe an individual document
Spreadsheet Metadata Fields Appropriate to Describe a Data file

Likewise, the above profiles could be applied heterogeniously though metadata attached to any level of the DSpace object hierarchy.

Metadata Field Inheritance

Individual Metadata Fields, like DCMI metadata properties will support subTyping or inheritance. For example, from the DCMI Website, we have the following:

Term Name:    title
Definition:A name given to the resource.
Type of Term:Property
Has Range:

Supporting a similar level of refinement for DSpace Metadata can be supported through the addition of new  MetadataFieldRegistry properties that are capable of storing this relationship.

Data Model Changes to Support This Proposal

To support this prpoposal, only additional fields and relational tables will be required to be added to the existing DSpace schema.


Profile will be used to identify a set of MetadataFieldProfile that define the fields allowed on a DSpace Object.


Individul field profile used to identify the basic rules allowed for a field assigned to a DSpace Object.


This table will be utilized to directly map any specified schema as a validation target for any existing DSpace Item. One, or more than one Schema assignment will be allow, creating a situation where an Item may be polymorphic and support more than one type.


This table will support the identification of which profile should be applied to new Items being created in any Collection within DSpace. This will be extended when support for metadata at all levels of DSpace is introduced, allowing assignment of Collection and Community "Types" to Community containers and likewise, support for Specific Bitstream types to be allowed in Item Containers.


A tentative list of new fields and tables is exemplified in the class diagram below.

The above solution can be easily encoded into the database schema, while the existing MetadataSchema, MetadataField and MetadataValue objects should be easy extendable to support new methods and business logic. 

A Example Use-Case

Metadata Schema Registry

In the following example an additional "dcterm" schema has been created to house the proper dcterms predicates while the "dc" schema continues to hold the existing qualified dc for legacy purposes.

Metadata Schema: "dcterms"

where "dcterms:xxx" refinements point to a new Schema in the repository that contains the fields required for the typical dcterms namespace.  In the current case, with the "item" and "item2" schema, this schema is not applied directly to Items, but inherited into defined "item" fields through "refinement".

IDFieldrefinesencodingdefaultrequiredScope Note
15dcterms.daterdf:PropertyW3CDTF${now}trueDate of publication or distribution.
25dcterms.identifierrdf:Property URI trueUniform Resource Identifier
37dcterms.languagerdf:Property RFC5646en Catch-all for non-ISO forms of the language of the item, accommodating harvested values.
40dcterms.relationrdf:Property URI   Catch-all for references to other related items.
57dcterms.subjectrdf:Property Literal   Uncontrolled index term.
64dcterms.titlerdf:Property Literal  trueTitle statement/title proper.
66dcterms.typerdf:Property Class  Nature or genre of content.

Metadata Profile Registry

The profile registry defines fields that may be attached to a DSpace Item.

  • A new "Profile" has been defined with its own namespace to be allowed on Collections A and B.
  • Each custom "Profile" can be applied to a specific DSO type (in this case, Item) via an "Applies To" mapping to objects that are of its type (in the diagram above, this is the profile2dso mapping).
  • Each custom "Profile" can enabled in a specific Container (Community, Collection, Item) via an "Allowed In" mapping  (in the diagram above, this is the profile2container mapping).

 IDNamespaceNameApplies ToAllowed In


Generic Item

ItemAll Collections
 2http://mydspace/schema/item2Simple ItemItemCollection A, Collection B

Item Metadata Profile "Generic Item"

The following exemplifies how a Profile for generic items that may have many optional fields attached to them.

elementrefinesencodingdefaultrequiredScope Note
issueddcterms:issuedW3CDTF${now}trueDate of publication or distribution.
datedcterms:dateW3CDTF${now} Use qualified form if possible.
uridcterms:identifierURI trueUniform Resource Identifier
identifierdcterms:identifierLiteral  Catch-all for unambiguous identifiers not defined by qualified form; use identifier.other for a known identifier common to a local collection instead of unqualified form.
isodcterms:languageRFC5646en Current ISO standard for language of intellectual content, including country codes (e.g. "en_US").
languagedcterms:languageRFC5646en Catch-all for non-ISO forms of the language of the item, accommodating harvested values.
haspartdcterms:relationURI   References physically or logically contained item.
relationdcterms:relationURI   Catch-all for references to other related items.
meshdcterms:subjectURI   MEdical Subject Headings
otherdcterms:subjectLiteral   Local controlled vocabulary; global vocabularies will receive specific qualifier.
subjectdcterms:subjectLiteral   Uncontrolled index term.
alternativedcterms:titleLiteral   Varying (or substitute) form of title proper appearing in item, e.g. abbreviation or translation
titledcterms:titleLiteral  trueTitle statement/title proper.
typedcterms:typeClass  Nature or genre of content.

Item Metadata Profile "Simple Item"

The second Item profile exemplifies a simple item with a smaller set of fields allowed, but with stricter requirements for populating those fields.

FieldrefinesencodingdefaultrequiredScope Note
issueddcterms:dateW3CDTF${now}trueDate of publication or distribution.
uridcterms:identifierURI trueUniform Resource Identifier
languagedcterms:languageRFC5646entrueCatch-all for non-ISO forms of the language of the item, accommodating harvested values.
meshdcterms:subjectURI  trueMEdical Subject Headings
titledcterms:titleLiteral  trueTitle statement/title proper.
typedcterms:typeClass trueNature or genre of content.

Steps To getting There

  • New fields and tables are added to database.
  • New attributes for existing DC schema and addition of the DCTerms Schema should be added to Registry after it has been extended.
  • Creation of several "Item Profiles" that can exemplify different types of Items in DSpace. each should utilize the new DCTERMS Schema where-ever possible.
  • Update DSpace build process to populate any necessary fields in new MetadataField and Profile tables.
  • Improve User interface and DSO data model to include returning details pertaining to Profile types for informing the User interface
  • Creation of new Describe Step and ItemEdit interfaces that enforce validation requirements expressed in the Metadata Profile
  • Creation of MetadataProfile Administrative Interfaces for managing Profiles.


The above proposal clarifies that new capabilities may emerge for "Typing" , "Restriction" and "Validation" of DSpace objects through extension of the existing data model.  The proposed strategy will support stronger typing of not only DSpaceObejcts, but also the values of metadata fields through validation rules such as syntax or vocabulary encodings, requiredness, Dublin Core or other metadata schema types.  DSpace should be able to utilize the new MetadataProfileRegistry as a means to replace large portions of the functionally found in the input-forms.xml file in future DSpace versions.

  • No labels


  1. I haven't yet read the whole proposal, I just have a comment to one part - the profile2dso and profile2container tables contain the handle_id attribute. They should contain resource_id instead (which can be looked up in the handle table to get handles). This is both to maintain orthogonality with other existing relations and to keep it easy to support other identifiers apart from handles in the future.

    1. The challenge is that if we do not utilize a more universal id such as the existing handle record, there is no ability to take advantage of foreign key constraints across all objects in profile2dso as some identifiers will refer to records in different tables.

      It is my belief that we need to work towards a database model where we can have one table containing a unique id for any DSpaceObject, the closest we have to this at this time is the handle table, likewise, in Richard Rodgers prototyping, he introduces a separate table to fill a similar role.  I am open to either approach, the ultimate objective being that we can start to simplify and generalize attachment of metadata to more than just DSpace Items.

      As such, fields like MetadataValueTable.item_id. would be shifted to MetadataValueTable.handle_id, and would become a foreign key representing any DSO in DSpace, allowing metadata to be attached to any DSO. Metadata Profiles would be created to enforce the fields required on specific DSO types.

      1. No argument from me against one identifier for all DSOs. Granted, the existing resource_type_id+resource_id is a composite key, so using that might complicate things (but would it?). And doesn't reusing handles as a unique internal identifier cause problems on its own, going against the direction of allowing multiple identifiers (everyone seems to agree we should support multiple identifiers) and simply being bad design (database table ids shouldn't have real-world meaning)?

        1. I realize this has been a year, but I wanted to just mention that the internal id's are specific to the "database tables" used by a specific DSpaceObject type, and that this really creates a limitation in extensibility of the DSpace model. To create a new DSpaceObject type, one needs to override the existing org.dspace.core.Constants class and add their own new integer ids for the new class. The tables that are really impacted by use of this integer are the resourcepolicy and handle tables. In an ideal world, there would be no need to store a integer based type id for the type of the DSpace object, the way this is achieved is not to map DSpaceObject types to integers, but instead have a common internal record for all DSpaceObject "Entities" at identifies the table name the object is stored in. Centralizing this relationship in an "Entity" table can be easily achieved by repurposing the handle table internally.  A more extensible model would look something like the following:


          CREATE TABLE Handle
          handle_id INTEGER PRIMARY KEY,
          handle VARCHAR(256) UNIQUE,
          resource_type_id INTEGER,
          resource_type_id VARCHAR(65535),
          resource_id INTEGER

          CREATE TABLE ResourcePolicy
          policy_id INTEGER PRIMARY KEY,
          resource_type_id INTEGER,
          resource_id INTEGER,
          resource_uid INTEGER REFERENCES Handle(handle_id),
          action_id INTEGER,
          eperson_id INTEGER REFERENCES EPerson(eperson_id),
          epersongroup_id INTEGER REFERENCES EPersonGroup(eperson_group_id),
          start_date DATE,
          end_date DATE,
          rpname VARCHAR(30),
          rptype VARCHAR(30),
          rpdescription VARCHAR(100)


          CREATE TABLE MetadataValue
          metadata_value_id INTEGER PRIMARY KEY DEFAULT NEXTVAL('metadatavalue_seq'),
          item_id INTEGER REFERENCES Item(item_id),
          resource_uid INTEGER REFERENCES Handle(handle_id),
          metadata_field_id INTEGER REFERENCES MetadataFieldRegistry(metadata_field_id),
          text_value TEXT,
          text_lang VARCHAR(24),
          place INTEGER,
          authority VARCHAR(100),
          confidence INTEGER DEFAULT -1

          Not only do you now have the ability to attach metadata to all DSpaceObjects, but you can extend DSO to create new types without needing a class hardcoding them all into existence.

          This of course requires that all DSpace resources have an internal uid identifier from the beginning of their existence, there is a hook for this already in the IdentifierServices, but it could be integrated into the existing DSpaceObject at creation time.  Finally, in this above example, we are repurposing the handle table for this, but if that is too conflated with the existing use of handles, it could easily be a separate table in DSpace.

          CREATE TABLE DSpaceObject
          uuid UUID UNIQUE,
          resource_type_id VARCHAR(65535),
          resource_id INTEGER

          CREATE TABLE ResourcePolicy
          resource_type_id INTEGER,
          resource_id INTEGER,
          resource_uid INTEGER REFERENCES DSpaceObject(id),


          CREATE TABLE MetadataValue

          item_id INTEGER REFERENCES Item(item_id),
          resource_uid INTEGER REFERENCES DSpaceObject(id),