Re: My task from last week: Semantic free identifiers

I tend to believe that "Perfect is the Enemy of Good".

It seems that everyone agrees at some point you need semantic identifiers.
 Look at the RDF, RDFS, and OWL standards.  They don't use non-semantic
identifiers (alphanumeric/guids) for core Classes and Predicates.  The
English-centric standard is an unfortunate defacto, because as far as I know
HTML, Java, etc reserved keywords aren't translated into different
languages, are they?  If so, its not the end of the world, we have
owl:sameAs/owl:equivalentClass/owl:equivalentProperty.

While it would be nice for the SPARQL standard to support alternative look
ups (aliasing using owl:sameAs?).  Lets say that somehow there was a core,
RDF tag that was put into SPARQL1.2 like "humanTerm".  But it wouldn't be
any different than it being a URI if it is supposed to be globally unique.

If instead we put the entire burden on the UI (and make the problem one
solvable by a product) we could have some sort of Code Complete/Label
Lookup, that you could start typing a label and it would offer suggestions?
 Nothing exists today, so perhaps it would be solved in a year or two with a
product.

If we differentiate end-users (who will have a GUI into the terminological
system and some sort of abstracted query capability) from "ontologists"
(that deal directly with concepts in the raw).  It doesn't make it easier
for ontologists to "deal with it with a tool".  You are relying on a tool
now to enforce a defacto relationship between a label and a URI.  At that
point, the two cease to be distinguishable anyway.

It seems like a great deal of work to enforce something that everyone wants
to circumvent anyway.  Or is not bothered by.

When applied to instances (but not to Classes and Predicates) I am less
adverse to having non-semantic identifiers.  Still, I haven't seen a really
good reason why that exceeds all the  reasons to use semantic identifiers.