So there's this Linked Data world where anyone can say anything about any subject, so one authority might say
dbpedia:Caves_of_Steel dcterms:creator dbpedia:Issac_Asimov .
but other people could say
dbpedia:Caves_of_Steel dcterms:creator fbase:/en/issac_asimov .
dbpedia:Cavas_of_Steel dcterms:creator basekb:m.03rx9 .
and we could probably find a bunch of others too -- these could either be stated explicitly or inferred through owl:sameAs.
So now I want to do a very easy SPARQL query like
SELECT * { dbpedia:Caves_of_Steel dcterms:creator ?creator }
and instead of getting back one answer, I get back a bunch of them. As a result, the user interface for my app shows "Creator: Issac Asimov, Issac Asimov, Issac Asimov" and the end user sees this and decides the semantic web won't be ready for prime time for another ten years.
Similarly, if I'm trying to convince people that SPARQL is simple and they can start using it, it doesn't help that they need to solve a PhD thesis problem just so they can process the results of a simple query like that.
So, if we grant that (1) it matters that people get the right answers to queries and (2) correct queries should be easy to write, how do we live with a world where U.N.A. doesn't hold?
How to live in a world, where there are hundreds of languages providing different names for the same thing, where there are millions of owl:sameAs links in the evolutionary growing Linked Open Data Cloud, and where the basic idea is that everyone can say everything about everything, if the UNA would hold?
My opinion: The UNA would be an immediate show killer for the web (of data)! I have never ever found a single good reason for the UNA, except for certain technical reasons, such as computational efficiency or stuff like that. In your case, I don't see a problem, and there would always be a way to do some sort of filtering solutions, if the result is not what one wants.
How To Live Without the Unique Name Assumption?
Consolidation/canonicalisation: choose one canonical URI for each group of coreferent URIs and, by default, only show canonical URIs to the user. If the user queries for a non-canonical URI, the system maps it to the canonical version in the background.
This is also used as an engineering optimisation by many owl:sameAs aware agents: choosing a canonical URI makes the materialised representation of equivalence linear as opposed to quadratic. In other words, taking the equivalent bunch of URIs U, instead of having to represent all pairs of owl:sameAs relations U × U and duplicating the data for each uri in U, selecting a canonical URI u' allows to compress the equivalence information into u' × U (∪ U × u').
Anyways, I agree with Michael. In the context of the Web, the more difficult question would be:
How could we Live With the Unique Name Assumpution?