Re: The UR* scheme registry, Citing URL/URI specs

Larry Masinter (masinter@parc.xerox.com)
Fri, 24 Oct 1997 11:46:25 PDT


Message-ID: <3450ED01.B1F06CB6@parc.xerox.com>
Date: Fri, 24 Oct 1997 11:46:25 PDT
From: Larry Masinter <masinter@parc.xerox.com>
To: Al Gilman <asgilman@access.digex.net>
CC: Dan Connolly <connolly@w3.org>, timbl@w3.org, fielding@ics.uci.edu,
Subject: Re: The UR* scheme registry, Citing URL/URI specs

I think that's a good summary of the situation. HTML and XML
can say they use URIs, and then point to a W3C note that
says "A URI is defined by IETF, currently it points to URLs,
and there is some work on URNs".

Here's a puzzle: 
A URN is a kind of a URL:
 One way to locate a resource is to name it. Of course,
 there's some assumption of a mechanism by which names
 can get looked up and turned into resource locations.
A URL is a kind of a URN:
 One way to name a resource is with its location;
 "the thing that is located at X". It's not a very
 good name, but then, not all names are good in that role.

Given both of these kinds of subsetting, it might be
that the distinction between "URL" and "URN" is not
in the protocol element (what's the syntax) but in the
use. URNs are used for identification, while URLs are
used for location.

What are the criteria for use of URLs and URNs?

Embedding in a document:

It's reasonable for a persistent form of a document to
contain its own URN, since the URN shouldn't change when
the document gets replicated. It is bad design to require
that a document contain its own URL (or its own BASE,
for that matter), because it interferes with replication
and transmission.

Document elements contain relative URLs, though, since 
a tree structure of URLs actually form a kind of compound
document, and the relative elements are used to identify
the other compound parts.

Larry
-- 
http://www.parc.xerox.com/masinter