W3C home > Mailing lists > Public > www-tag@w3.org > January 2003

Re: yet another sidetrack on what a URI identifies

From: Roy T. Fielding <fielding@apache.org>
Date: Tue, 14 Jan 2003 19:12:16 -0800
Cc: www-tag@w3.org
To: Sandro Hawke <sandro@w3.org>
Message-Id: <26CCED8D-2837-11D7-9D22-000393753936@apache.org>

> So what does the URI "http://www.w3.org/Consortium/" denote?   If you
> can't say for certain, then tell me what it might or should denote.

Whatever www.w3.org wants it to denote.  What do the RDF assertions
say it denotes?  That is, after all, why we needed RDF -- because the
Web doesn't tell the client what a URI denotes, nor what properties
it will maintain consistently over time.

If I were to use it as a link and perform a GET, then I would expect
the response to consistently contain information about the W3C as
a consortium.  Note that I said the RESPONSE will contain information,
not that the resource IS information.  Given those semantics are
relatively stable, I could then use the same identifier as a
replacement anywhere that I might refer to the "W3C organization".
Am I right in doing so?  That entirely depends on the owners of
www.w3.org -- maybe that's not what they intended by the name.

Does "http://www.w3.org/Consortium/" denote the web page I just
received?  No, but my cache does identify that page as a representation
of that resource at the time it was generated by www.w3.org, and thus
my cache has stored it as such with the URI as the cache key and
metadata about the representation to indicate its freshness.

My point is that the URI alone is not sufficient to target assertions
unless you assume method="GET" and time="for all time", which seems
a bit too limiting to me.  Or, you could claim that what you are
building is an abstract KMS consisting of nothing but nodes and
typed relationships, and in fact that all of these URIs are just names
and have no relation to the information system of the Web.  In that
case, there are no Web pages, no information to be transferred,
no operational hypermedia that needs modeling, no caches, and no
pesky users to get in the way.

In any case, the reason we had this discussion originally is because
some people were complaining about xmlns identifiers being http URIs
because they believed that a URI could not be both a name and a way
of retrieving a web page.  They are wrong, as demonstrated repeatedly
by working practice and the REST model, because they were ignoring
the difference between a URI and a GET action on a URI.

....Roy
Received on Tuesday, 14 January 2003 22:13:12 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:15 GMT