URI dereferencing and authoritative information ("tag" scheme versus "http" scheme)

Charles McCathieNevile wrote to W3C’s Semantic-Web discussion list
(<mailto:semantic-web@w3.org>) on 9 October 2005 in “Re: New Intro to
RDF” (<mid:op.syebyxeewxe0ny@widsith.local>,
<http://www.w3.org/mid/op.syebyxeewxe0ny@widsith.local>)

> [“tag” and “urn” URIs] fail even [to identify resources with no Web-
> accessible representation] to the extent that to find out what 
> [those URIs] represent [people tend to] look at the web.

This is not a failure of identification. For the sole purpose of
identification, a “tag” URI is as good as an “http” URI. Both can serve
as unique names. For description, an “http” URI with the backing of a
good HTTP server with the backing of a persistent organization is the
clear winner. Let us distinguish identification from description.

The problem is knowing which information is authoritative.

Presume that the Internet host “example.org”, through an HTTP service on
port 80, is the definitionally authoritative source for information on
the resource <http://example.org/>.

Presume a temporary misconfiguration of the HTTP service such that the
response to any request for <http://example.org/> bears a status code of
“404” (not found). During the misconfiguration, what is the
authoritative information on <http://example.org/>?

Presume a surreptitious take-over of the host “example.org”, including
deliberate contradiction of previous information. Presume an indefinite
duration for the take-over. During the take-over, what is the
authoritative information on <http://example.org/>? Does the URI
"http://example.org/" change meanings?

On the other hand, if an “http” URI’s resource has no definitionally
authoritative information, then “http” URIs are no better than “tag”
URIs for description.

As long as the default channels of information remain open to subversion
and decay, the dictum to “follow your nose” is merely a heuristic. What
seems necessary is a web of trust and explicit statements about sources
of information.

-- 
Etan Wexler.

Received on Saturday, 15 October 2005 07:21:12 UTC