RE: [PORT] Subject Indicators

Alistair,

* Steve Pepper

| > I just want to give you a heads-up that my current
| > thinking is that having a special property may not be
| > the optimal way of adding subject indicators (and hence
| > Published Subjects) to RDF. I am tending toward the idea
| > that a *class* defining the concept of "information
| > resource" (as a subclass of "resource") may be a better
| > way to go.

* Alistair Miles

| Am most interested to hear your ideas on this.
| 
| Without knowing any more detail about your current thinking than what you
| said above: the idea of defining the class of 'information resources' sounds
| like a good idea ... but that doesn't address the issue of how to identify
| non-inforesources.  Do you think it is OK to allocate URIs directly to
| non-inforesources?  Or do you think we should always identify
| non-inforesources indirectly (e.g. as described in [1])?  
| 
| ..
| 
| [1] http://www.w3.org/2001/tag/webarch/#indirect-identification

This is interesting.

Re-reading [1] it turns out to be thoroughly unclear (at least
to me) what is meant by "indirect identification".

That term can be understood in two completely different
ways, which I describe below as senses (1) and (2).

Indirect identification (1)
...........................

   Identification via a URI which does not resolve directly
   to the subject being identified.

This is the sense in which I have used similar terms in the
past,[2] and which I find to be the more useful of the two.

The thinking behind this use of the term is as follows:

ONLY network addressable information resources (otherwise
known as "documents") can be identified DIRECTLY (via their
locator or address).

EVERYTHING ELSE (including most of what the semantic web
wants to identify) can by definition ONLY be identified
INDIRECTLY, because most things don't exist as strings
of bits and bytes on the web.

In other words, Nadia, the person (cf. [1]), CANNOT by
definition be identified DIRECTLY.

Given this position, your question "Do you think we should
always identify non-inforesources indirectly [...]?" makes
no sense: My response would be: WE HAVE NO CHOICE! We can
ONLY EVER identify non-inforesources indirectly.

But, as I said, on re-reading [1] I see that the notion of
"indirect identification" could be interpreted differently:
i.e., as describing the MECHANISM by which a URI used as
an identifier becomes a "characteristic" of an RDF node
representing the subject in question. (I'm using the term
"characteristic" in order to avoid having to say either
"property" or "attribute", which would have sufficed if
they hadn't had more specific meanings...)

Thus, the value of the [about] attribute on an
<rdf:description> element could be considered a DIRECT
identifier, whereas the value of any property whose purpose
is to provide an identifier (such as skos:subjectIndicator)
could be considered an INDIRECT identifier. This leads to
a second definition of "indirect identification", quite
different from the first:

Indirect identification (2)
...........................

   Identification via a URI that is the value of a
   subsidiary property (as opposed to the URI that is the
   "built-in" or "intrinsic" identifier of an RDF node,
   which is used for _direct_identification_).

[Struggling a bit with the terminology here, not being
an RDF maven myself, but I hope the point is clear.]

If (2) is the sense in which you pose your question, my
answer is: I do indeed think it is OK -- even preferable,
and clearly in line with much current practice -- "to
allocate URIs directly to non-inforesources".*

HOWEVER, having said that, what I also strongly advocate
is that those URIs follow the requirements of the OASIS
Published Subjects TC, the most important of which is
that the URI should resolve to a human-interpretable
subject indicator.

For that reason, I find the introduction of the property
skos:subjectIndicator well-intentioned but ultimately
counter-productive since it will encourage people to assign
"direct identifiers" (in the sense (2), above) that DO NOT
resolve to subject indicators.

Best regards,

Steve

* This would be "direct identification" in sense (2) and
  "indirect identification" in sense (1).

[2] http://www.ontopia.net/topicmaps/materials/identitycrisis.html#N246

--
Steve Pepper <pepper@ontopia.net>
Chief Strategy Officer, Ontopia
Convenor, ISO/IEC JTC 1/SC 34/WG 3
Editor, XTM (XML Topic Maps 1.0)

Received on Thursday, 2 December 2004 09:46:03 UTC