The difficulty I see in this thread (and much prior to this thread) is that instead of contending with a single approach to bibliographic description couched in the implementation language of the time (book, card, and MARC catalog technology) we now see two approaches where workers are trying to do theory in implementation (or at least IT-centric ) languages.
Not all of the issues associated with Cultural Heritage resource description fall with the scope of IT-centric descriptive languages.
Early 20th century approaches to scientific description (observers, observations, reference frames, etc.) and for representing systems of relationships (graph, node/vertex, link/edge, subgraph, degree centrality, etc.) together (along with conceptual contributions from the social sciences) provide more than enough – implementable – descriptive power to address library/archive-grade resource description requirements.
You have to apply the two approaches simultaneously.
Let me add a few points to that:
* We need to take into consideration the viewpoints of people who are experienced in both sides of the current effort (bibliographic description and linked data), and there are very few people with a great deal of expertise in both.
* Actually, in my view, having people engaging with the effort that do NOT know MARC is an advantage, not a disadvantage. There are a huge number of peculiarities introduced over the years that should be reconsidered by fresh eyes, rather than perpetuated without cause.
* There's more future than there is past. MARC has existed for a long time, and there's a lot of legacy there. But there's even more bibliographic descriptions that have yet to be written. The longer we remain tied to this horribly outdated technology, the more legacy data we will need to deal with and the further apart from the rest of the world (that is part of the web, not just displayed via the web) we become.
Your mileage may vary :)