Allow meta-data about nineml-spec

[apdavison]

I vote for using RDF. We could make recommendations for which ontologies to use (e.g. Dublin core + the computational neuroscience ontology) but I think any RDF-formatted metadata should be valid.

[tclose]

I don't have a strong preference with the exception that it is not too complex

[tclose]

NineML standards meeting:

It was decided that for version 1 any NineML element can contain an <Annotation> sub-element. Any valid xml can be put in the annotation element, however it is recommended to use a standard annotation markup language such as RDF.

[apdavison]

NineML standards meeting:

It was further decided that to be labelled as "NineML compliant", any tool handling NineML descriptions must preserve all existing annotations, except where a user explicitly edits/deletes them. This needs to go in the spec.

[tclose]

I was just thinking whether should we make it mandatory to use RDF? It would probably greatly improve its value if all NineML files were using the same system for annotations wouldn't it?

[ajc158]

I'm happy to use RDF as a requirement - we just need to run through some use cases and check that RDF is going to be a sensible and intuitive way to apply them. For simple descriptions it clearly works, we just need to think out things like extension data to check it is the best option.

[tclose]

That sounds good, although we probably don't need to use it for everything, only the things it makes sense for.

[tclose]

What do we need to do in order to support this issue for version 1.0?

[apdavison]

I think we just need to make sure that the Python library preserves annotations when reading then re-writing a model.

[tclose]

PR #60 should implement this for the Python library.