-----Original Message----- From: McElwain, Paul Benjamin [mailto:[log in to unmask]] In my working on the Variations FRBR implementation, as a data modeler, I was struck by little attention had been paid to the relationships by the FRBR Report. I'm not surprised though, treating the relationships at at "entity" level (having their own attribution) is a more obtuse exercise of abstraction. We do treat relationships as attributed "entities", for more information about the "role" involved. (a creator may be as a composer) One way to think about the originality of an expression of a work, being the first ever expression, could be as an attribute of the relationship between the work and expression. Paul... **** Paul, I agree that from a theoretical perspective it makes a lot of sense to model the original expression as an attribute of the relationship between the work and that expression. Or to do what FRBRoo did and make classes like Work Conception and F28 Expression Creation. It's not really clear to me that in our particular situation there is any practical advantage to trying to do that rather than creating a merged work/primary expression entity. This has to do with the kind of expressions we're mostly modeling and the way we're trying to model them. Most of the moving image expressions that average libraries deal with are defined not by what I think of as bundled attributes, but are rather a set of independent attributes. This is unlike the typical music expression, which I think of as a set of bundled attributes. If you know you have a performance of X work on Y date in Z venue then, if someone has previously created in an expression record, you know a number of other things about that expression such as the composer, performers, and arrangement of the piece without having to re-verify them again. All those things could productively be stored as a unit. This also happens with film with various cuts such as airplane versions or director's cuts. These do have some associated attributes, notably length, but also perhaps a different editor. For the kinds of unbundled attributes that are common with moving images, especially DVDs, there are a large number of attributes, like soundtrack and subtitle languages, accessibility options (captions, audio descriptions), and aspect ratio, that vary independently. With these kinds of unbundled expression attributes, a cataloger has to reexamine all of them every time there is a new manifestation. If there's a change in our knowledge of what subtitles are on a specific manifestation, it does not have automatic implications for any other manifestation that might have that same constellation of options. The other types of attributes that describe the original expression of a film are those that never change because they are important facts about the history of the work that we want to note in conjunction with any future expression. Many of these are things that RDA says are attributes of expressions that moving image catalogers would tend to think of as attributes of works (e.g., casts and costume designers do not vary among expressions so why record them on every expression?). In a sense, at least for moving images, the original expression is bit of an abstraction and in practice we get most of our information from reference sources. At first, I thought we could just model these unbundled attributes of the expression as attributes of the manifestation/publication since, as I mentioned above, they have to be verified with every new manifestation anyway. Work record 1 Dracula (1931) Tod Browning English Manifestation record 1 1 VHS videocassette (1985) OCLC#: 13754402 Audio: English I ran into trouble with manifestations that include more than one work. Some still work well enough, either because the expression-level information is all the same or is unknown. Work record 1 Ursula (1961) Lloyd Michael Williams English Manifestation record 1 1 DVD video (2005) Experiments in terror ISBN: 0976523922 Audio: English Work record 2 Journey into the Unknown (2002 ) Kerry Laitala English However, in some cases, the expression-level information varies between two works on a single manifestation/publication. The manifestation below includes two versions of Dracula, each in its original language. For the prototype, I just made two different manifestation records, which repeat most of the same information. That doesn't seem to me to be a desirable long-term solution. Work record 1 Dracula (1931) Tod Browning English Manifestation record 1 1 DVD video (1999) ISBN: 0783227450 Audio: English Subtitles: English or French Work record 2 Dracula (1931) George Medford Spanish Manifestation record 2 1 DVD video (1999) ISBN: 0783227450 Audio: Spanish Subtitles: English or French So I think we do need the intermediate expression level, but I am not sure if we need an actual expression record as in: Work record 1 Dracula (1931) Tod Browning English Expression record 1 Audio: English Subtitles: English or French Manifestation record 1 1 DVD video (1999) ISBN: 0783227450 Work record 2 Dracula (1931) George Medford Spanish Expression record 2 Audio: Spanish Subtitles: English or French Or just independent statements directly linking the work and the manifestation: Work record 1 Dracula (1931) Tod Browning English Audio: English Manifestation record 1 1 DVD video (1999) ISBN: 0783227450 Subtitles: English Subtitles: French Work record 2 Dracula (1931) George Medford Spanish Audio: Spanish Subtitles: English Subtitles: French I'm not sure if I explained this very clearly. I have trouble getting my own head around it sometimes. But FWIW, this is some of the thinking behind why we did what we did the way we did it. It may in fact be a very film-specific approach. Kelley