Clearly, all names are not going to be established to NACO standards. If I am understanding this correctly, there is an alternative between establishing everything in a formal, vetted vocabulary like NACO and just using strings. You could to give something an identifier using a local domain (one such as mit.edu/person/5005). Or perhaps some organization like OCLC could provide a shared space for unvetted identifiers.
I think this what what Karen Coyle mentioned as the intent of Bibframe's lightweight abstraction layer (which I will have to look up again). It sounds like you would connect your work to a new local identifier and attach the string label (authorized access point) to the local identifier instead of directly to the work. It is then easier later to map that local identifier to a NACO one if it becomes available or to some other local identifier than it would be if you just had the string attached to the work. It sounds like there are a lot of practical details to work out, though.
You could also think of this as similar to local authority records that some libraries make even if they don't contribute them to NACO. I do agree with the comment about the symbiotic relationship between strings and authorities in that you also need to know what the name looks like on the resource.
Kelley (who is still thinking through all the other messages on this thread)
From: Bibliographic Framework Transition Initiative Forum [mailto:[log in to unmask]] On Behalf Of Benjamin A Abrahamse
I totally agree with this. Any future encoding standard needs to support both "strings" AND "things" and not force cataloging institutions to use one to the exclusion of the other.
For example, my library is part of NACO but we don't participate in BIBCO, and so we won't establish every name we run across that doesn't appear in NAF. Only if it's (a) an MIT-related person or corporate body; or (b) a name that will cause a conflict in our author file, will we definitely establish the name. (We will also establish things for a variety of other reasons that fall under the rubric of "cataloger's judgment", but these are the two situations where authority work is, a priori, required.)
If we were forced to operate in an environment where we did not have this flexibility it would definitely have a deleterious affect our productivity.
Moreover I think there is a symbiotic relationship between libraries who record strings, and the libraries (sometimes, the same library at a different period of time) who will come along later and turn them into authorities. You sometimes need to see a name 2 or 3 times in the bibliographic file before you have enough data to establish a person with any certainty.
Just like it is important to record (as a string) what you see listed as the publisher, and not simply grab an "authorized form" of the publisher's name and slot it in. If there is no one recording these changes in the real world, there will be no data available to document what the variants or changes might be for the authority record.