RE: getting the ball rolling (or at least providing a target to shoot at)

Trying to catch up with the debate ... I basically agree with Karl
viewpoint, and particularly on this:

> I think the meaning of URI, but I may be wrong, is more constrained by
> the context and for what applications they are used but more by the URI
> itself. Intended meaning and understood meaning are things which can
> have a lot of different shapes.

There are basically two stakeholders in the use of a URI. Let me call them
hereafter "Publisher" and "User".

"Publisher" is to be understood in the Dublin Core definition : "An entity
responsible for making the resource available". Note that in that case, I
understand resource as the URI in its context of publication (namespace,
DNS ...) Publisher is IMO easier to define than the notion of URI "owner"
that has been previously discussed, and is a widely used and understood
concept.

"User" can be whatever from inference engine to data base application to
human user interface (browser or otherwise).

Of course there is a context of publication, and there is a context of use.
As Karl points out, using a URI is always a negociation process between
those two actors in their respective contexts, and the meaning a result of
this negociation. So we can set the issue as following:

- How can Publisher explicit the context of publication and the
possible/recommended/mandatory context(s) of use for the URIs he/she/it
publishes.

- For each context of use, how the meaning must/should/may be
endorsed/redefined/refined by User.

As Peter remarked, current best practices in the (Semantic) Web can provide
good starting points. I should quote here the path opened by Topic Maps use
of URIs, making the distinction between two contexts of use as either
"subject address" or "subject indicator". See for example Steve Pepper's
"Curing the Web Identity Crisis" at
http://www.ontopia.net/topicmaps/materials/identitycrisis.html.

IMO this is a good starting point, but certainly too simplistic. It takes
into account only a given context of use (Topic Maps), and not the context
of publication. We have tried in OASIS Published Subjects TC to work out
what should be done on Publisher side for facilitation of such a context of
use. I must say that after two years of work, we did not came to very
substantial conclusions. But maybe we were focused on a too specific type
of negociation.

I have also tried with some success in pragmatic applications with Mondeca
software, the notion that an URI can be used to identify basically three
types of things : information resources, real-world identified individuals,
or abstract concepts (universals), and that each of those uses triggers a
specific behaviour of the application. Without specification of the above
quoted negociation and context, almost any URI can be used for any type.
But it's quite easy to specify the context in applications. If for example
http://www.w3.org/People/Connolly/ is used as an identifier of an
individual instance of "Person" in my data base, it will be clearly have a
different meaning (read : trigger a different behaviour) than if it's used
as an identifier for an individual instance of "Personal Home Page". Both
make sense, but I can't used both is the same context without risk of
inconsistency or weird behaviour.

Now the Publisher of the above URI (W3C or Dan himself, amazing I could not
figure exactly from the heavy semantic markup of that resource) could
specify in which context(s) this URI can/should/may be used, as a basis for
negociation with User.

Of course this example is maybe too easy. But there are many others as easy
or even easier, and mostly those URIs which are really intended by
Publisher to mean something - and among those the URIs identifying formal
classes, individuals and properties in OWL. So should not we focus on those
cases where there is a meaning declared on Publisher's side, and let alone
the cases where there is not (in those latter cases, let User set the
meaning, with all the inherent risks).

Thanks for your attention

Bernard Vatant
Senior Consultant
Knowledge Engineering
Mondeca - www.mondeca.com
bernard.vatant@mondeca.com

Received on Wednesday, 24 September 2003 08:48:37 UTC