I think the thrust of your argument is bang on the nail. It chimes strongly with work I did some time ago (and am building on further currently). http://events.linkeddata.org/ldow2008/slides/RobStyles_SemanticMarc.pdf

There is a key difference in the semantics of data found in MARC records and the data we would like to publish as Linked Data. The data in the marc record is a statement of what is printed on the book, not a statement of truth.

So, where it says “Publisher Statement” that’s because it’s the statement made in the book, which all comes back to book-in-hand cataloging for the purpose of stock management and discovery within a library.

This is key because the statement printed in the book will not change over time, whereas names and locations of publishers change as companies merge, split, go broke and re-form.

What is needed is both - literal values for what is printed on the book-in-hand (hence tying it to the manifestation in most cases) but also properties referring to the organizations and peoples involved. We can start to build on data mining techniques and bringing in external data to populate those properties.

Thanks for prompting more thought about this.

rob