Re: Ontologies versioning?

On Mon, 23 Jan 2017 08:34:46 -0800, Chris Mungall <cjmungall@lbl.gov> wrote:
> Yes, most OBO ontologies derive the version from the ISO-8601 of the 
> date. This is used in the versionIRI, e.g. 
> /obo/releases/YYYY-MM-DD/<artifact> and often in things like github 
> releases tags. See for example:
> 
> https://github.com/EnvironmentOntology/envo/releases
> 
> I like semver for software and can see it working well for schema-type 
> ontologies and upper ontologies.

Agreed, ISO8601-dates makes more sense for database-like knowledge ontologies
(where our understanding of the real world can change under out feet!), and
semver-versions for schema-like ontologies.

It might still make sense for knowledge ontologies to indicate a major version
that corresponds to which schema it is using (in a rough sense) - e.g. an
ontologies describing species could have:

 owl:versionIRI "2.20160529.0"

which could be replaced by a newer:

 owl:versionIRI "3.20160602.0"

Here 3.x indicates that even though the ontology is just a few days older, it
is incompatible in its structure to the former - e.g. your old queries will
most likely break - say it is now using a different upper ontology or schema to
describe relationships between species.  Moving :Dolphin from :SeaCreatures to
:Mammals however would just be an update to the minor version.


(This still complies with semantic versioning, btw - I would keep the patch
version starting with .0, so you can increment it when you 'broke some tiny
thing' in packaging the dated release and need to "re-release".  You know it
can happen!)


-- 
Stian Soiland-Reyes
University of Manchester
http://www.esciencelab.org.uk/
http://orcid.org/0000-0001-9842-9718

Received on Thursday, 26 January 2017 11:16:18 UTC