Re: [URN] Re: URI documents

Roy T. Fielding (fielding@kiwi.ics.uci.edu)
Fri, 02 Jan 1998 23:40:26 -0800


To: Larry Masinter <masinter@parc.xerox.com>
cc: Patrik Faltstrom <paf@swip.net>, harald.t.alvestrand@uninett.no,
Subject: Re: [URN] Re: URI documents 
In-reply-to: Your message of "Sat, 27 Dec 1997 22:45:15 PST."
             <34A5F57A.F02C06D7@parc.xerox.com> 
Date: Fri, 02 Jan 1998 23:40:26 -0800
From: "Roy T. Fielding" <fielding@kiwi.ics.uci.edu>
Message-ID:  <9801022347.aa29408@paris.ics.uci.edu>

In message <34A5F57A.F02C06D7@parc.xerox.com>, Larry Masinter writes:
>I've been wondering if we might progress with a different
>perspective, which I will lay out briefly:
>
>The term "URI" denotes a concept: some kind of uniform
>space of identifiers for resources. The terms "URL", and "URN",
>however, denote explicit protocol elements, with specific
>syntax and semantics associated with that syntax. The URL
>syntax is well defined and widely implemented; the URN syntax
>is newer, but well underway.
>
>It is possible, and convenient, to allow for mutual embedding
>of these protocol elements. That is, the space of URNs can
>be embedded in the space of URLs by prepending the URN with
>the string "urn:". This means that "all URNs are URLs" only
>in the trivial sense that it is possible to write a URN as
>a URL. The simple embedding is possible because the URN syntax
>was chosen (wisely) to fit within the URL syntax.
>
>Conversely, one could imagine embedding the URL name space
>in the URN space. Imagine, for example, a URN which consisted
>of a date and a URL. The resource named would be 'the
>resource that was available at the specified location
>at the specified time. This would be perfectly good URN;
>perhaps one could say that 'dated-url' was the naming authority,
>and write <durl:199712272238:http://www.parc.xerox.com/masinter>.
>as the undecorated URN.
>
>With this notion of 'URI is a concept', we don't need or want
>a general URI syntax. There's a URL syntax and a URN syntax.
>There might be a completely separate URC syntax at some point.
>While it is convenient to embed URNs into the URL name space,
>it is just an embedding; the semantics associated with URLs
>are not inherited by URNs than they are inherited by email
>addresses merely because there is a 'mailto:' URL scheme.

This is what we were working on, until the URN folks insisted that
a URN could not be a URL, and therefore could not be referenced
by the URL specification.  Likewise, many people feel that
"identifier" is a better term for this interchangeable syntax
than is "locator".  If people could just make up their minds on
which is better 

   o URN can be considered a URL for some purposes

   o URN can be considered a URI for some purposes

then maybe we can make some progress.  Please note that we have already
written both of those specifications, minus a few adjustments needed to
correspond with more recent editorial improvements.

....Roy