Yeah, what you describe as a “manifestation” and “edition” (or what FRBR would call an “item” and “manifestation”) are pretty straightforward. Likewise, we can generally agree that there’s a top-level “work” concept that can be used to group some set of “artistically identical” editions/manifestations under a single-heading. The questions then require some understanding of what constitutes “artistically identical” and how, exactly, BookBrainz would intend to define and break up works:
- Is a work a conceptual entity consisting of a single creative burst by one or more individuals, however expressed or transformed? (i.e. do translations, illustrated and critical editions, etc. belong to the same work?)
- Is a work a single artistic product defined by the significant creative forces that combined to produce it, but not transformative ones? (i.e. are illustrated editions distinct works because the illustrator is different, but translations and critical editions might be the same work)
- Is a work an artistic product defined as the result of all significant creative contributions and transformations that have the potential to significantly alter the meaning of the work? (i.e. are translations, too, distinct works)
Some of these distinctions can, of course, be captured by intermediate levels. For example, it might be reasonable to say that the Edition Group concept represents some single “artistic edition” (i.e. grouped by translation, or by illustrator, or critical commentary, or textual revision, or both), while a Work represents the abstract entity. Indeed, I suspect this is sort of what the Edition Group was intended to solve.
Likewise, relationships in the data model can implicitly express some of these groupings without an explicit entity in the data model, like how MusicBrainz can relate works as being “translations” of other works. (Of course I’m not suggesting that this is necessarily appropriate for BookBrainz, since translations of works are not nearly as common in music as they are in literature, which suggests that a higher-order grouping of such translations might be more useful)
Of course here I still glaze over other subtleties:
- What about works that are embedded in journals, anthologies, and other collected forms? (i.e. where there is not a one-to-one relationship between Work and Edition)
- For that matter, what about introductory material in an edition that is ancillary to the primary work? Does it constitute a second, independent work? Or is it just implied by the Edition (Group)? (And if the latter, do those introductions mean you have no way to obviously group Edition Groups that only differ by the introductory material?)
- What about works where the author revisits and revises a portion of the work with artistic (rather than strictly grammatical/editorial) intent to supersede a previous edition, like Tolkien’s second edition of The Hobbit?
- What, if anything, does the anthology, journal or collected form have as its representation, if not a work? Certainly it can have multiple editions and revisions, as well.
- Do we even want to start talking about how comic books are released and compiled? Every issue of Neil Gaiman’s The Sandman has been pressed and collected in at least 5 different forms (original serial, trade paperback, Absolute Edition [which saw some minor artistic revisions and extra content], Annotated Edition [with textual annotations], and Omnibus Edition [re-collected Absolute Edition content into two, rather than five, volumes]) and that’s a series that probably doesn’t start touching on the modern(?) concept of variant covers of individual issues that otherwise contain the exact same internal content. I’m sure other comic books are equally confusing to sort through.
Sorry to be posing so many questions. I don’t think these all need to be answered now, but I do think they are all worth considering as the model evolves, especially because the existing data will need to eventually be rectified to fit any changed model.