Re: My task from last week: Semantic free identifiers

I think all those who suggested "semantic-less" or "meaning-less"
identifiers should come out and define what they mean by
"semantics/meaning".

Does a URI "http://example.com/foo#good" carries the semantics of *good*
as how the word "good" is defined in an English dictionary? Unless there
is a specification that binds a (or part of a) URI to natural language,
then a URI is semantics-less or meaningless at the first place.

The argument that an identifier, with part of it looking like a natural
language word, can be stale is just absurd. If the URI
"http://example.com/foo#protein_x" is used to refer to a biological
entity, say x, can be stale because we later may find that x is not a
protein as we previously thought, then the identifier
"http://example.com/foo#12345" can be stale for the same reason.

What make them stale is not that the URI contains a substring "protein".
Rather, it is because the assertions we have made of it, such as,

http://example.com/foo#protein_x a bio:Protein.
http://example.com/foo#protein_x rdf:comments "protein_x is a protein ...".

Are those who favored the so-called semantic-less identifiers suggest that
if we use "http://example.com/foo#12345" instead, then, we can just leave
all the other assertions unchanged?

If we think an identifier like "12345" is "meaningless" but "protein_x" is
meaningful, it is only because we don't know or accustomed to how the
former is used. A lot of planets/stars, for instance, are named with
"meaningless" string (at least to most biologists). But, should the
astronomers invent another set of "meaningless" identifiers for the set
they are already in use?


Xiaoshu Wang

Received on Tuesday, 21 June 2011 17:12:02 UTC