URIsandNamespaces

It's important to disassociate the concept of URI from namespaces. I think that it is important that a certain set of instances curated by a particular group are named in a way that helps us to identify this. Whether we like it or not, URIs have evolved from naming documents on the web, people expect to get useful information from a URI, and even expect to type it or some subpart of it into a web browser and get back something. What I would could the subdomain part of a URI, ie. the uri up to and not including the local unique ID for an instance(or some other rdf resource) gives an identifier for a subdomain of modelling that is under some particular groups curation. They are responsible for any objects referenced using this subdomain. They may for instance have a specific policy on freshness of resources, or policies about complaining about resources. The RDF community make various noises to this effect, and at the end of the day, OWL is just extended RDF and any OWL document is readable by RDF tools - in fact I only use RDF reasoners at the moment. So my vote would go to each database group building a URI for an instance model based on their own web domain plus some useful path.

As for rules to create the unique parts of the URI, there are many. One nice way for simple database records is to use a hash, such as an MD5 hash of the content of the instance - say the property values, or the entire structure. That way it is easy to check if the resource at the end of the pointer is actually the right one. However, this changes the URI for the smallest of changes in the data, which can render lots of references to the old one useless in other documents. So I don't think that is a good model where there is a lot of cross-referencing, unless the cross-referencing is at the class level. It certainly wouldn't work for CellML. Another way is to introduce major/minor versions into the path that represent various degrees of change in the data. This is a model that would suit CellML for instance, such as making enough changes to a model that the mathematics and behaviour of the model has changed.

The process for interpreting multiple databases would be to simply import the different URIs of instance models from different database curators.

As for default and base namespace. I'd leave that as simply an OWL/XML issue. They choose what they want to. If you really think about it, then it only makes sense to use the namespace from which the instances are based. So that would imply most people simply use the namespace of the model they are creating instances of. In those cases where they extend the model then they would use that namespace.