- Open freely accessible content available via universal repository (federated union of many indiviual OAIs)
- Authors self contribute content items and the metadata describing them. They maintain ownership, but license for the free use of others (Creative Commons).
- Metadata is at a miminum Dublin Core, and should adhere to standard controlled vocabularies (ontologies).
- Classification/Ontologies of relationships between concepts should be relational based (multi-faceted), not hierachial (just like databases evolved).
- Content items have permanent universally unique IDs (DOIs, ISBN) *as well as* a URL (the "handle" indicating how to retrieve it). Authors have permanent universally unique IDs.
- Comments, indexings, annotations, are full class citizens as content items. Not something attached to "original content item" in a subordinate manner. All information is equal, and stored as the same type of unit.
- Search interface is one stop shopping as entry point (Google), with ability to refine search based on metadata. Interactive refinement of search.
- While initial discovery of content items (or containing collections) is done through Dublin Core metadata (or possiblly fulltext/fullcontent) searching, a second stage of searching may occur utilizing specific application interfaces for that data type. For instance locating a particular section of video within a video collection using a video browser.
- GOAL: a user should be able to find items matching small easily defined search terms within seconds, or a larger set of less homogenous items within a few minutes at most.