OwlVersioning

Introduction

Here you will find documentation on the distributed versioning implementation of OWL ontologies. This should familiar to any programmer with experience with a DVCS (http://en.wikipedia.org/wiki/Distributed_revision_control).

We do have one generalization of how things are usually done in the DVCS word in that we want to support what one might call composite versioning where one create an aggregate of several versioned objects and tracks versions of the aggregate as a separate entity. More on that below.

Concepts

The ontology is set of axioms. Modifications to an ontology amount to adding and removing axioms from that set. Or adding/removing imports or prefixes or annotations, which are extra-logical things that are simple structured elements nevertheless. A list of such modifications is a ChangeSet. Note that even though we call it a set it's actually a list. That's because the order of changes matters. Given an ontology in some initial set, one can apply a ChangeSet to create a new Revision, which essentially represents a version of the versioned entity. The term revision is used in version management because a revision also represents the act of creating a version. So the word "revision" has this double meaning of sorts in this context.

Revisions are connected to each in a parent-child relationship. They form a DAG (directed acyclic graph).

Change sets also form a DAG, but because of composition they don't necessarily result in a revision. For example, when a creating a new revision of a project comprised of several modules, each module will have a change set committed, but it will be the project that will acquire a new revision, not the modules. Modules can be versioned independently.