It's hard for me to imagine that those using linked library data will be putting what is retrieved from the network directly in front of their patrons without intervention. That's not really technically feasible, whatever the desire may be.
It’s not even really desirable to present the raw data directly to most users in most cases, anymore than it would be to present them with a raw MARC record (looking the tags and offsets in the directory is a real pain). Assembling the entities from streams of RDF triples is almost as painful to do by eye. :)
We will need to be guided by the Web architecture and use a design with caching. If you cache remote linked data resources locally (and if you intend to give your patrons a reasonable experience, you will be caching) you can certainly make emendations into or out of the cache, processing data in whatever ways you see fit.
Exactly! Adding assertions about some Entity is one of the things that Linked Data makes really easy. If an IRI denoting some fictional thing lacks the appropriate characterization, you can make the appropriate assertion yourself; for example adding an assertion that the entity is a member of the class of FictionalThings. This can be published for others to use.
A well designed future bibliographic network for open data would allow for all assertions about a given entity to be retrieved cheaply; indeed it is probably feasible to copies of all assertions locally using simple overlay techniques. Data can be fused from multiple sources.
Commercial cataloging providers might require a subscription to access their data, and strictly license its use and reuse.
Simon
