Just some short notes (and questions) to one of your (very valid and interesting!) questions:
> Some time ago there was a discussion about recording the place of
> publication. Places seem to me to be an example of the best case
> for using URIs: there are comprehensive, externally-maintained
> lists at the level of specificity required (generally cities for places of
> publication). In most cases, it seems like just using a URI and
> abandoning transcription for place of publication would be
> functional. This wouldn't work for things like early printed books,
This is probably true if there is only one place mentioned on the title page, but what do you do with something like "Berlin ; Heidelberg ; New York ; Hong Kong ; London ; Milan ; Paris ; Tokyo : Springer, 2004) ? Are all of those really places of publication? In that respect early printed books are easier, because a publication statement like "published by Georg Moony in the High Street and printed by P. Albert Murgatroyd in the Lower Crescent at the Sign of the Lantern"  definitely refers to people (who can have URIs) who had offices and workshops that are so locateable that you probably could supply GPS coordinates for them.
> but for most contemporary materials, it would seem sufficient to
> make a note if there were something unusual about the way the
> name was presented on the resource (and does BF have a way to
> connect notes to the elements that they are describing?). From a
> practical perspective, the place of publication still needs to be based
> on the resource rather than being a characteristic of the publisher.
> Publishers move around and have offices in many places. Trying to
Yes. Nonetheless I think that an authority file with publishers would be an excellent asset (it's just a specific case of corporate bodies...).
> track who was based where and when would be a nightmare. Using
> URIs would have the benefit of distinguishing London, England from
> London, Ontario and London, Ohio.
Yes, that is definitely an advantage!
 made up by me, but I hope it conveys the point...
> Names are another area where we will make use of URIs.
> There are a great many more names than places and, although
> there are multiple external sources of URIs, there is nothing
> like the comprehensive coverage that is available for places.
> What happens if there isn't an existing URI for a name? There is
> a cost to making a string into a useful thing and putting it in its
> place in the universe. The cost is lessened by eliminating the
> need to create a unique string, but it is not zero. Not everyone
> will be able to contribute to shared lists like the LC National
> Authority File. Not every name is worth the trouble of
I guess this is a general case: What do I do when I want link to an authority file and the entity I want to link to isn't there (yet)?
> Will people just coin a one-off URI using their own domain? How
> reliable will these be? The advantage of this approach is that if
> new information becomes available, it's easier to integrate. You
> can just say that this locally-maintained URI represents the same
> thing as this NAF identifier and not have to mess with the string.
> If you start out with the assumption of separate until proven the
> same, this might work reasonably well since it's easier to merge
> than to split apart.
It probably depends on how your library data ecosystem is designed. I could imagine that when your cataloguing system finds out that the entity you are looking for does not yet exist (neither in your system, nor in the central authority file) it creates a temporary entity in your system and enters that into a workflow to have it added to the central AF. Once it's there, it updates the references in your system.
My two cent (€),