Re: URIs / URLs

"Charles F. Munat" wrote:
> Some types of resources on the Web are volatile by nature. John Doe puts up
> a personal web page. Three weeks later, he decides he doesn't really want a
> personal Web page - too much trouble to update it - and deletes it.
> It seems to me that there are some resources that *should* be persistent,
> others that have no need to be. Has anyone given any thought to how to
> distinguish the two? Should RDF mark persistent documents as such? Should
> anything not so marked be presumed to be volatile? Are there different
> degrees of persistence (e.g., a page that is persistent while some of the
> data on the page changes, or a page that represents the latest version of a
> document - but always that document).
> What about the way in which these documents are referenced. Would it be
> better to have both permanent and temporary links? Could links be set to
> expire? Could documents include an expire date and links to that document
> automatically expire when the document did?

Obviously, stability and longevity are relative and subjective terms. Some
would consider one year, some five, some one hundred as a stable time period.
This varies by individual, community, and application. Rather than try to 
navigate that particular thorn patch, XTM [1] leaves this up to the author. 

If as a topic map author I am trying to create identifiers that I want 
people to use as such, I advertise them not simply as URIs but as Published
Subject Indicators (PSIs) [2], which means that I intend them as stable 
indicators. I might publish a list of PSIs in a specification, an online
or print trade journal. These identifiers could be used in my community
and would have whatever stability I claim. Life is unpredictable, but
differentiating a URI as a PSI means that I intend them as identifiers.
Obviously, my publication of a PSI is different than say, the US Library
of Congress. But that's what we call "competition in the marketplace", eh?

In my work with the Cyc ontology [3] one of the first things I endeavoured
to do was ascertain how stable the HTML documentation for Cyc 2.1 was, and
contacted Doug Lenat in this regard. Upon his reassurance I published a
set of "canonical" URIs that can be used as PSIs to indicate Cyc constants
in the upper ontology. These have been published as a list [4] and can be 
used today. Neither I nor Doug will change the list for Cyc version 2.1.

We may at some point in the future need to topic-map the relations between
the current list and some newer version, but the point I'm trying to make
is that absent any true legal contract one *never* has a guarantee of
stability. That doesn't inhibit the use of URIs as identifiers, unless
one expects a specific result upon resolution., the W3C and
even IBM might not be around in ten years, but the identifiers still will
identify the same things as the day they were first published, and are
thus useful.

Your question addresses more specifically resources, and you correctly
point out that there's little that can be done to guarantee their 
stability. You touch upon many of the outstanding challenges in IT, and 
I'm encouraged by the ability of technologies such as RDF and XTM to be
able to represent things like confidence levels, expirations dates, etc.
outside of the resources themselves. These problems don't go away via
RDF or XTM, but certainly seem more manageable simply because there is
a standardized methodology to provide metadata about resources and now 
map it into existing ontologies and topologies.


[3] [BTW, this is very
    much a work-in-progress]
Murray Altheim, SGML/XML Grease Monkey     <mailto:altheim&#64;>
XML Technology Center
Sun Microsystems, 1601 Willow Rd., MS UMPK17-102, Menlo Park, CA 94025

      the wood louse sits on a splinter and sings to the rising sap
      ain't it awful how winter lingers in springtimes lap -- archy

Received on Tuesday, 10 April 2001 18:12:44 UTC