another crack at 'information resource'

Another attempt at an 'information resource' theory.

(I mean 'information resource' in a sense close to TimBL's, not
in David's sense.)

As I said earlier, if you assume you're doing design (architecture)
and not ontology (what exists), the problem gets easier, since you're not
constrained by reality, truth, etc.; you don't have to be able to
"identify" information resources in the sense of pointing your
finger at them, or falsify the "is a representation of" relation.

Rather than try to figure out what IRs are, let's try talking about what we
would like to say about them, i.e. properties that they're supposed to
have, and then if the question remains, speculate on what we'd like
them to be like in general.

Many of the properties that one would like to assert are about
content: author, title, date written or published, subject matter.
These properties can vary from one observer to the next, or
through time: the author, etc. might vary, so we have to decide how
the properties of the IR relate to the properties of its
'representations.'

It doesn't do much good to assert a content property if it can't be
corroborated by someone else observing the IR [credit Larry
Masinter], therefore I propose the
principle that a content property is true of an information resource
iff it's true of *all* of the resource's representations.  E.g.

<http://www.w3.org/TR/2010/WD-rdfa-core-20100422/> dc:creator :Ben_Adida.

The quantification is time-bounded in the same way any RDF statement
is, thus reducing the time problem to one that's previously unsolved.

A "fixed resource" in TimBL's sense of the word would be an IR that
shares in all the content properties of its representation.

An annoyance: Suppose that Fred is the author of some
representations of Z, and that Fred is not an author of
some other representations of Z. We can infer that Fred
is not an author of Z (assuming "is not an author of" is not
considered a content property).  - This will surprise some people, but
the alternatives are nasty.  Existential instead of universal
quantification would very quickly lead to inconsistencies.
Having the implication go only one way is too inferentially weak to be useful.
Having no connection between representation properties and
resource properties gets us back to total representation/resource divorce.

There's another set of properties having to with ways the IR *does* vary,
e.g. http://news.google.com/ is a news site (representations are
lawfully related to current events), the weather in Oaxaca example from AWWW,
http://en.wikipedia.org/wiki/Special:Random ,
http://www.w3.org/TR/rdfa-core/ (related in a particular way to a
social process overseen by W3C), and so on.  The content
properties of representations vary, but they do so in a regular way.

This leads me to the suggestion that an IR is (or is induced by, or is
pretty much the same as, or is characterized by, or is a theory of) a
lawful regularity among representations.  (Compare computer science
'invariant'.  Also similar to API.) I don't see any way to make them
more definite than that without tying them to their incarnations in
ways we've been trying to avoid.  They are inventions, theories, phlogiston,
monads. Like FRBR works or expressions only worse.

Maybe IRs need a few more conditions.  I would think that physical
realizability is a condition, i.e. if it can't be put on the web at
least in principle, it doesn't qualify.  And of course our two
conditions "no physical properties" and "not mathematical" -
IRs are like other abstractions such as promises and songs in
this regard, so we ought to be able to make those conditions stick.

Here's something else I wrote today related to this:
http://odontomachus.wordpress.com/2010/11/08/to-what-does-a-uri-refer/

Jonathan

Received on Monday, 8 November 2010 17:51:04 UTC