Content type creation (in our case, MgdSchema) could be made more user friendly by using ontology design patterns. That way when a content creator wants to create a new content type (say, conference) they can derive it from a common pattern in a centralized repository (say, event).
How should CMSs react to different contexts? Could there be generic ways to handle things like user location, device used to access the site, languages and other context information.
Midgard has context injectors but those are only the API hook for this, there still could be some conventions for this.
There are many ways to extract semantic information from documents:
Named entity recognition ("this is a person's name")
Relationship analysis ("this is a feature of a product")
Content analysis ("this table looks like an ingredient list")
After this kind of data has been generated from document, it could then be microformatized, or for instance added the the RDF triple store of the document. And then made available for semantic searches ("recipes containing this ingredient").
The IKS stack could provide various tools for extracting and processing semantic information. These tools obviously need a way to access the content in the CMS. Common interfaces like JCR or CMIS could be used for this.
Looking at the list of open source CMSs involved makes me wonder what unique advantages each system provides so that it needed to be built, considered that there have been pretty good CMSs available since late 90s.
On the other hand, the multitude also highlights the need for interoperability and common guidelines.
