Databases Reference
In-Depth Information
5.2
Summary
Table 6.3 shows a comparison of most systems that are discussed. While the first
two systems Protege and KAON support complete processes for ontology evolu-
tion, OntoView and OnEX focus on the management of existing ontology versions
developed elsewhere. Supported ontology formats are primarily RDF and OWL;
Protege and OnEX can integrate further formats (e.g., OBO). With the exception
of OntoView, all systems support both simple and complex changes. The represen-
tation and determination of an evolution mapping between two ontology versions
differs among the systems. Protege is most flexible for specifying ontology changes
by supporting both incremental changes and the provision of a new ontology ver-
sion; the other systems follow only one of the two possibilities. A Diff computation
is supported by all systems except KAON. The update propagation to instances
and related data is partially supported in KAON and OnEX. KAON uses evolution
strategies to adapt instances managed together with the ontology. OnEX supports the
identification and migration of annotations affected by ontology changes. With the
exception of KAON, all systems support sequential versioning. Graphical user inter-
faces are provided by all systems: Protege and KAON are editor-like applications,
while OntoView and OnEX are web-based.
6
Conclusions
Effective schema evolution is a long-standing problem that is difficult to address
since schema changes impact existing instances, index and storage structures as well
as applications, and other schema consumers. We introduced the main requirements
for effective schema evolution and provided an overview about the current state of
the art on the evolution of relational schemas, XML schemas, and ontologies. More
than 20 approaches have been analyzed against the introduced requirements and we
used several tables to compare most of these approaches side by side. The introduced
methodology should be similarly applicable to evaluate further schema or ontology
evolution approaches. We summarize some of our observations as follows.
Commercial DBMS currently restrict their support for evolving relational sch-
emas to simple incremental changes and instance migration, while there is not yet
support to semi-automatically propagate changes to dependent schemas, mappings,
and applications. Filling this gap requires support for the determination and pro-
cessing of expressive schema mappings that have been studied in recent research
approaches such as Pantha Rei/Prism and in model management research (
Bernstein
and Melnik
2007
).
The evolution of XML schemas is easier than for relational schemas since the
schemas can be extended by optional components that do not invalidate exist-
ing instances. Due to the absence of a standard schema modification language,
schema changes are usually specified by providing a new version of the schema.
In research approaches, schema matching and mapping techniques are being used
