Thanks so much for the information, but I am looking at it another way. In my experience, when you are working with any tool, there are different ways of working with it: the way it was designed to work, or not. People who are experienced will use the tool as it was designed to be used; those who are not experienced use that same tool in different ways.
So, if you are using a power saw, someone with experience will use it in one way, while someone without experience will use that same power saw completely differently, probably incorrectly, and may saw off a body part or two. The experienced person may figure out some innovative ways of using that power saw, but there are still certain ways of using it that cannot be transgressed.
This is the way I view the library catalog: it is a tool, and its basic design occurred in an evolutionary manner over a long period of time. But, it has not changed in its particulars since it was codified primarily in the mid-19th century. The primary means of control are through the use of "consistency" in the creation of bibliographical description, and by the use of "organization" in the creation of what we always termed "access points," i.e. the formal places in the catalog where searchers could find a card (or an entry in a printed catalog). Therefore, if you decided not to make a card for a title with author main entry (which happened), that means that it was *impossible* to find e.g. Twain's Huck Finn under the title, unless a cataloger made a mistake.
It is only the organization of the cards in the catalog that make a heading coherent. For example, only the organization of the cards make a heading such as "Agnelli family--Homes and haunts--Italy--Villar Perosa--Pictorial works" less ludicrous than it seems when taken out of the catalog. When it is removed from the catalog as I did here, people will say, and I agree, that *nobody in the world* would ever think of anything in such terms. "Homes and haunts"???? "Pictorial works"???? But, when seen in relation to the other "cards" around it, the heading begins to make sense. See: http://catalog.loc.gov/cgi-bin/Pwebrecon.cgi?DB=local&Search_Arg=agnelli%20family&Search_Code=SUBJ_&CNT=100&hist=1 and do some browsing. You will see how the headings work together, with various subheadings for the family, then comes each family member, often with their own subdivisions, and so on and so on.
There is a power shown here in the catalog that is not found in search engines. The experienced librarian understands how this tool is designed to work, but fewer and fewer patrons today know this; many fewer understand it, and their numbers are diminishing all the time.
I apologize for this lengthy discussion, but I fear that this sort of knowledge is being lost. For centuries, the premise of how our tool worked was based on browsing, because that was all anybody could do. The introduction of keyword eliminated this. The modern view is that the library catalog is a database, and while I agree with this I add the proviso: it is a card catalog stuck into a database, where a lot of things that worked fairly well before don't work very well at all now. These basic premises were never seriously rethought and catalogers continue to make headings that are designed to be browsed.
But we certainly agree about opening up the data for experimentation. This is the only solution. I think that the headings we make, even the crazy-looking ones, are still very useful, but something must change somehow. Making APIs is definitely the way to go, but that essentially means that the records will "be taken out" of the catalog into a new environment. As I tried to show, the result will be that the headings will look ludicrous, and this is not even mentioning how people can go about finding them. We must find solutions.
These are some of the tremendous changes of the possibly not-too-distant future that we need to be dealing with. And why I sort of think that discussing what are manifestations, items, expressions and works is a bit like sweeping the porch when the tsunami is coming.