As someone who once spent a great deal of time in retrospectively populating bibliographic records of the same title (records for frbr:manifestion associated with the same frbr:work) with consistent subject headings, the significant appeal of FRBR’s WEMI structure is the potential for inheritance of such relationships down the WEMI tree (cascade as Stephen puts it).
With apologies for being contrarian, if BibFrame is not going to support such inheritance, but instead will require the repetition of identical relationships in separate BibFrame:Work records representing frbr:expression entities in the same frbr:work family, then it seems hardly be worth the investment being poured into BibFrame -- this implied functionality of the inherent WEMI relationships under FRBR was one of the few labor offsets to be found in the model, and a huge boon in terms of reliably and consistently populating subject terminology throughout a WEMI tree, both at the inception of the tree and for subsequent maintenance in the event subject terminology refined at a later point.
I will go further to state that BibFrame, as articulated from its earliest presentation, has shortsightedly been shackled far too tightly to the Bibliographic vs. Authority record dichotomy we have in our current MARC environment. The articulation of an expanded set of attributes and relationships for frbr:Works, frbr:Expressions, frad:Person, frad:Corporate_Body, and frad:Family; the subsequent expansion of corresponding elements in RDA; and the further subsequent expansion of fields in the MARC21 formats (accompanied by numerous instances of tensions between formats to record similar information) all indicate that all of the entities put forth in the Functional Requirements (FR) family of models need to be fully articulated in the next generation communication framework, and be done so on a level playing field. The differentiation between Instances and Works, and between Instances and Authorities, along with the implicit privileging of Instance at the center of the BibFrame model is significantly problematic in this regard, as is the merger of frbr:work and frbr:expression under the BibFrame:work.
We are hamstrung in MARC from re-articulating a structure of equivalent entities, since the LDR/06 byte that holds record type has had too many of the available alphabetic characters consumed in designating various bibliographic format record types (while all the authority records are glommed under ‘z’). While there are some few vestigial structural differences indicated by the current breakdown of bibliographic types, it is a questionable arrangement post-format integration. A new structure could entail four for Group 1 types, three for Group 2 types, a minimum of four for Group 3 types (subject to prospective expansion of the Group 3 types as noised about in some quarters), and an unknown number for various genre aspects that remain unaddressed in the FR Family but are being actively explored and developed by an ALA/ALCTS/CaMMS/SAC working group. With only eight characters left unused in MARC21, we would have to break the format to realize this -- more's the pity, but we would not be so constricted in a new framework such as BibFrame is offering.
There was a discussion paper presented to MARBI a year or so back addressing the designation of bibliographic (and possibly authority) records for use as any of the WEMI entity “types.” There were significant structural issues with respect to expanding employment of the formats along such flexible lines, so the proposal was dropped. But it was my sense also that NDMSO didn’t “get” the data architecture being proposed. If that sense was accurate, it appears to be symptomatic in how BibFrame has been developed.
It should be possible for BibFrame to be conceived and implemented at a higher level of generality, while still accommodating legacy data as a special restricted subset with its dichotomous bibliographic/authority structure. Gordon Dunsire was able to provide a preliminary reconciliation of the widely divergent FRSAD model with the previous two FRBR and FRAD models, by articulating the latter two as subsets of the former (don't ask me how: FRSAD makes my head hurt, but I understood Dunsire enough at the time to see that it is possible). If such a reconciliation and nesting of concepts can be done in that context, it can be done for MARC in a BibFrame environment. And BibFrame would be well served by looking at the FR reconciliation process so as not to be caught a generation behind at the starting gun.
Schaffer Library, Union College
Schenectady NY 12308
Frankly, we've not really addressed this (though we're aware of the idea of inheritance in this sense). It’s not the we won't, it's more to do with seeing where the data goes and what is practical.
The nice thing - as I see it - about BIBFRAME Works that double as RDA/FRBR Expressions is that, when the information is repeated, the BIBFRAME Work can stand alone without reference to another BIBFRAME Work (what would be the RDA/FRBR Expression). Mind you - it's not that there is no link to a BIBFRAME Work that is representative of an RDA/FRBR Work (there is), it's just that you do not also need that other BIBFRAME Work to make sense of the one that is representative of the RDA/FRBR Expression.