W3C home > Mailing lists > Public > uri@w3.org > September 2002

Re: Possible new issue: Things with and without identity?

From: Al Gilman <asgilman@iamdigex.net>
Date: Tue, 10 Sep 2002 09:40:45 -0400
Message-Id: <>
To: Stefan Eissing <stefan.eissing@greenbytes.de>, "Roy T. Fielding" <fielding@apache.org>
Cc: Miles Sabin <miles@milessabin.com>, uri@w3.org

At 08:09 AM 2002-09-10, Stefan Eissing wrote:

>Am Dienstag den, 10. September 2002, um 11:43, schrieb Roy T. Fielding:
>>On Monday, September 9, 2002, at 08:44  PM, Miles Sabin wrote:
>>>A discussion point from the rest-discuss list which is probably better
>>>dealt with here,
>>>   http://groups.yahoo.com/group/rest-discuss/message/2465
>>>At issue is the first sentence of the informal definition of resource in
>>>RFC 2396 1.1,
>>>   A resource can be anything that has identity.
>>>"that has identity" is redundant because *everything* has identity in
>>>the only reasonably straightforward understanding of identity, ie. the
>>>logical truth in all but the most obscure formal systems that,
>>>   (Vx) x = x
>>No, that is not the only reasonably straightforward understanding of
>>identity.  That is the specific mathematical axiom of identity and not
>>identity itself.  We already had that discussion on www-tag.
>>Everything does not have identity.  "Everything" includes both things that
>>have been identified and those that have not yet been identified.
>>only becomes a resource after it has been identified, but not necessarily
>>by a URI.
>>The definition is correct as it stands.
>Maybe as an add-on to this:
>WebDAV working group decided to call a URL without a resource
>(e.g. 404 on GET) an "unmapped URL".
>In that context, one could call a resource without URI a
>"unmapped resource".
>So instead of going metaphysical with names and identity, it
>maybe makes more sense of talking about *the* mapping
>between the set of resources R and the set of names N
>and define properties of this mapping M.

Here you have assumed the point that I was attempting to
correct.  You assume that the category of resources are
a point set, and that URIs operate within this point set.

URIs operate within a set of communicative gestures we
term 'reference' and not all of them deal in the identity
properties of things, or in persistently entified 'things'
at all.

>M: R -> N
>M is neither injective, nor surjective, that is
>   there are (many ;) x in R without n in N so that M(x) = n
>   there are n in N without x in R, so that M(x) = n
>and as stated before: |R| > |N|
>Of interest to RFC 2396 is the subset of names which are uniform
>and follow a specific syntax. Let's call the set of such names
>U. So what RFC 2396 describes is which elements of N are
>also elements of U. Elements of U are called URI.

This all sounds nice, but it it unreal.  URIs are not all names.

URIs are used (search and data: URLs) to refer by characteristics that do
not qualify as names.

In the particular case of a search URL, there is no persistent thing.  The
'representation of the resource' received is a thing but the 'resource
identified in the request' is not.

The things that persist are the information model with regard to which the
query is written and the language for formulating a query against that
model.  And the service of synthesizing a thing that conforms to the type or
category specification defined by the query received from the client.

This distinction is not metaphysical, it has to deal with the structure of
computational processes linked by the references encoded in URIs.

>"Having identity" means than that there is a mapping between
>U and R. Such a mapping can be established in several defined
>ways and, as of the uniformity, in endless, future, not-yet
>defined ways.
>For the subset of URIs "below" http://www.greenbytes.de, my
>web server is the authority over the mapping between those
>URIs and the set of resources on my server (which might no all
>have a URI as it might refuse to offer any
>file://greenbytes.de mappings or might not even have a file system).
>(If I take my server down, the mapping with URI is lost, although
>the resources still exist and probably have names given to them
>by the operating system.)
>So, if we restrict RFC 2396 to only talk about the mapping of
>URI to resource and defining that "having identity" in RFC 2396
>means that such a mapping exists (established by whomever),
>do we then get rid of the metaphysical issues?

You do the web a disservice to dismiss the architectural issue as
'metaphysical.'  It is very computational and economic and hence bears
importantly on how we frame our axioms for the Web.

The social process of denomination is not sustainably scalable to where the
web is going, because it is

a) costly
b) unnecessary

If we don't understand that URIs run on (derive their value in service from)
their use in the calculus of description, and not particularly a naming
calculus, we are going to leave the bulk of the future Web 'dark' by our
lights and wind up leading our own little backwater while the main stream of
the Web goes elsewhere following other pipers.

To understand the operating conditions of the Web one has to view it in
terms of relative unity, integrity and persistence and not view these
characteristics as boolean-valued.  Because at the actual operating point of
the web, "what URIs refer to" is characterized by fractional relative unity,
integrity and persistence.  This is what the market will bear.  This
operating point is sufficient to suport persistent expectation that
following URI-references will add value in the client user's world.  The
fact that this expectation is satisfied, in the main, keeps customers coming
back.  But only in the main.  It doesn't need to get perfect, and it isn't
about to do so.


Received on Tuesday, 10 September 2002 09:41:15 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:04 UTC