Page History

...

The table doesn't model the metadata fields (e.g. dc.title, dc.relation.uri, ...) as column names (called attributes in the relation model), but uses the EAV (Entity-attribute-value) model instead, modeling the field name itself as a value in the metadata_field_id column.

The EAV model is typically used for modeling sparse matrices (tables where there is a large ratio of NULL cells to cells containing data). But the reason why it is used in DSpace is that it makes changing the metadata schema (adding/removing fields) easier to do in the application (DSpace) rather than the database (where DDL must be used, which may take a long time and typically makes the table unavailable until the operation is completed). Adding/removing metadata fields is thus effectively a routine user operation (available to what DSpace calls the "Administrators" group) rather than a task for the sysadmin of the DSpace server.

...