Someone on the RDA-L list posted a question, that if I understood it right, was asking, of a given example of FRBR modelling, “Why is the movie version of Pride and Prejudice considered a different work than the novel?” (And, incidentally, the shooting script for the movie is considered yet another work again).
This is a choice, but the recommended/consensus modelling choice right now is indeed to call the film version of Pride and Prejudice a different (creative) ‘work’ than the novel, and the film script yet different again.
This is a somewhat arbitrary choice — when modelling reality, we have to make choices on how to ‘summarize’ reality in our modelled data, in the most useful ways for our use cases. It is my opinion that neither choice is neccesarily more ‘right’, any model is neccesarily a summarized ‘lossy encoding’ of reality. The menu is not the meal, the map is not the territory, the model is not reality.
In this case, however, that choice is arguably/probably the most consistent with legacy cataloging practice, where a film version gets a different authority record than the original novel — now that alone maybe isn’t clear “evidence” because authority records are sometimes at the ‘expression level’.
But probably more importantly, the film version also gets a different ‘main entry’. Things that are the same ‘work’ in legacy cataloging practice are going to have the same main entry, if they have different main entries, that means legacy cataloging practice treated them as different works.
Sort of, it’s a bit ambiguous — we’re kind of reduced to looking for ‘evidence’ of modelling intent in legacy cataloging (from both the existing data and the existing rules), because it was not explicitly, clearly, or formally defined anywhere — the existing rules, despite Lubetzky’s attempt to be “principles based”, are really just step-by-step recipes, they don’t tell you the intent, or explicitly lay out the modelling decisions. (The model is NOT reality, so you have to describe the way you model reality).
Part of the point of FRBR/RDA is to make it less ambiguous and more consistent, but (for better or for worse), follow the lead of our inherited legacy practice. The FRBR report, and RDA too, quite clearly are trying to build on legacy practice, formalizing and it making it more explicit and more internally consistent, but not changing it wholesale.
Some people like FRBR/RDA for this, some people hate it for this. I tend to think it was the right choice — don’t try to do too much at once, first formalize and make explicit and consistent what we’ve got, and THEN we know we’re all talking about the same thing to consider changes.
Additionally, a 150+ year modelling tradition (I suggest that’s what ‘cataloging’ is, modelling) that developed incrementally probably got a lot of things right, don’t change it’s modelling decisions too hastily. (Analogy: Rewrite your whole software from scratch is going to end up beingway more of a pain than incrementally refactoring, you lose a lot of “encoded knowledge” represented in the source code, the legacy of a bunch of bug-fixes and such.)
And on top of that, we’ve got a giant legacy ‘system’ (the records, the software, the people), when building a new system, we’ve got to try to be mostly backwards compatible. Being able to turn existing authority records into mostly-basically-right Expression and/or Work records is important, if you wholesale change the modelling decisions (“oh, the film is now part of the same work as the novel!”), it limits the possibilities of making any sense of legacy data using new formalized model.
(On the other hand, the FRBR/RDA attempt to make those legacy practices and unspoken ‘modelling choices’ more explit, formalized, and internally consistent, is hugely important. It’s just that trying to do that and simultaneously completely rejigger the implicit modelling choices of cataloging… whew, it’s biting off enough without trying to do that, even if the rejiggering would be an improvement.)
So, anyway, the modelling choices say that a novel and a film based on it belong to different ‘work’ sets — but they can certainly still be related by OTHER relationships, such as a work-to-work relationship “is based upon”. Don’t make the mistake of thinking the FRBR model says two things are either part of the same work, or they’re not related at all — the FRBR model encourages a variety of relationships between things.