W3C home > Mailing lists > Public > www-tag@w3.org > October 2007

RE: Resources and representations (was RE: Subgroup to handle semantics of HTTP etc?)

From: Williams, Stuart (HP Labs, Bristol) <skw@hp.com>
Date: Wed, 24 Oct 2007 11:07:00 +0100
Message-ID: <C4B3FB61F7970A4391A5C10BAA1C3F0DEE048B@sdcexc04.emea.cpqcorp.net>
To: <wangxiao@musc.edu>
Cc: "Booth, David (HP Software - Boston)" <dbooth@hp.com>, "W3C-TAG Group WG" <www-tag@w3.org>, "Alan Ruttenberg" <alanruttenberg@gmail.com>, "Jonathan A Rees" <jar@mumble.net>, "Dan Connolly" <connolly@w3.org>, "Tim Berners-Lee" <timbl@w3.org>

Hello Xiaoshu,
<snip/>

> > However, I guess what I'm objecting to is that you seem to me to
slip 
> > into speaking of the representation that you obtained from a given 
> > resource as being that resource.
> >   
>
> Yes, that is the point that I was trying to make.  The 
> bit-stream manipulated by my browser is not the thing 
> identified by the URI, it is a representation of it. 

In that case we are agreeing, probably loudly on this point, and when
you give me cause to think otherwise I'll assume that I have misread you
and move on.

> At 
> most, it is a copy of it but this is hardly to know just from 
> the representation alone.


<snip/> 

> >> That is where the problem is.  A URI denotes a thing but
> >> *never* the description of that thing.  The *representation*
returned 
> >> by the server is a description of that thing.
> >>     
> >
> > The way I have come to square this is that there resources which are

> > descriptive in nature eg. some RDF description of a thing (or
things), 
> > or some HTML description of a thing (or things). That description
(or
> > depiction) is a separate resource from the thing described and 
> > deserving of its own URI. You can GET representations of a
descriptions (or a
> > depicition) but those are *not* representations of the
described/depicted thing.
> >   
> Then there are two kinds of things now.  One is 
> *description*, the other, let's call it, *subject*.  You have 
> one URI, you choose one to denote either the *description* or 
> the *subject*.  You cannot denote both, right? 

There are two things, yes. A thing and a description of a thing (which
is also a thing).

There are also *TWO* URI one for each.

If you're using # URIs then you arrange things thus:

  http://example.com/things#dog		refers to a dog;
  http:///example.com/things			refers to a resource
that includes (maybe amongst other things) a description the referenced
dog.

The request line of the HTTP protocol does NOT carry the fragment part
of the URI.
You do not, cannot, get a representation of the #dog URI by doing HTTP
get using that URI.
What you get is a representation of something else... using David Booths
terminology,
you get a representation of whatever the radix of the URI refers to (or
at least you try to).

> > Now a question arises (and we have been there, along this path - 
> > search for Partick Sticklers dog or Dan's car) as to whether a 
> > representation of description can also serve as a representation of 
> > the described thing. The TAG's httpRange-14 resolution does advise 
> > *not* to do that in the case of things such as people, places, 
> > celestial bodies, physical artifacts... use either '#' based client 
> > side indirection or 303 protocol indirection to a resource that 
> > describes (possibly amongst other things) the thing of interest.
> >   
> The httpRange-14 is asked on a wrong assumption that a URI 
> can denote both a subject and its description. But, it cannot. 

The HTTP range question simply asks what sort of things can an HTTP URI
refer to?
And the answer given is 'any kind of thing' (whether or not their is a
'#' in the spelling of the URI).

> >> When people think that
> >> "http://www.ihmc.us/users/phayes/PatHayes" should NOT denote Pat 
> >> Hayes because they have confused the returned
> >> *representation* as the *resource* denoted by that URI.
> >>     
> >
> > No... its because they have confused the descriptive 
> > document(resource) that the representation is a representation of
with 
> > the stated referent
> > - or at least find the state of affairs inconsistent.
> >   
> Why they get confused the 
> "http://www.ihmc.us/users/phayes/PatHayes" as descriptive 
> document? Who said that? Pat certainly does not.  is it 
> because its *representation* is parsed in an HTML browser 
> instead of a hologram?

What is a machine to make of it?


[Aside: despite the declaration within (I guess) Pat :-), a great many
references are made to
http://www.ihmc.us/users/phayes/PatHayes.html about which the narrative
is silent :-)]

> What I was trying to say in the past 
> few days are, you cannot infer the nature of *a resource* 
> from the nature of *its representation*.  You must read the 
> message and see what it says.

Largely I do not disagree. 

> >>> b) leave the range of what can be referenced by an http: URI sans 
> >>> fragment, unconstrained.
> >>>   
> >>>       
> >> This becomes non-issue if we straighten out (a), yes?
> >>     
> >
> > Well... it becomes a non-issue for those that regard a
representation 
> > of a depiction of a thing as an adequate representation of that
thing 
> > (for whom a representation of a depiction of a dog is an adequate 
> > representation of that dog). But there are several around for whom 
> > that is *not* the case.
> >
> > Careful assignment of URIs can be arranged so that on can speak 
> > separately about say, a dog, and its depiction (for example they 
> > likely have different dc:creators).
> >   
> The word *adequate* just goes back to the *essential 
> characteristics*, I hope we all drop this once for all. We 
> cannot tell objectively which case is adequate and which are 
> not.  Let's drop them and treat every resource in the same way.

Well... over the course of this multi-year debate I have come to regard
as wrong to deploy resource representations for people, places, planets,
concepts etc. These are things that admit description or depiction but
IMO at least defy representation (in the sense of
webarch:Representation). But a thing and its description (of which there
can be many and conflicting) are on the whole distinct (there are some
things that might be regarded as self descriptive). So... assign them
distinct URIs, and for things that defy representations (yes that's a
judgement call) don't deploy webarch:representations, but if you want to
be helpful organise your URIs so that someone 'enquiring' about the
thing *knowingly* obtains a (one of potentially many) description.

<snip/>

> > c) is 303 or '#'.
> >   
> If we have a unified view, this is no longer issue. URI is 
> URI slash or #.  All behave the same.

>From a denotational POV yes. From a techical access POV, no.

<snip/>

> A representation is by nature descriptive, right?  Why do I need a
> *representation* of a *descriptive thing* to describe a thing 
> at the first place?

When you retrieve from:
http://www.nasa.gov/worldbook/jupiter_worldbook.html

What does the representation describe?
What does the representation represent?


> > <snip/>
> >   
> >>>> Why cannot. Assuming your past few email is posted to my website,

> >>>> wouldn't it possibly change the state of my mind?
> >>>>     
> >>>>         
> >>> But... can you convey your "state of mind" before and after our 
> >>> exchange in a message? :-)
> >>>   
> >>>       
> >> But isn't this just a matter of implementation detail?
> >>     
> >
> > I don't know... is the current state of you mind expressible in a
> > message?
> >   
> Of course not all of it but part of it, can't you tell?

No! :-)

> Regards,
> 
> Xiaoshu

I have a feeling that we have become repetitive and that we should stop
(at least onlist). I think that there is considerable common ground, but
also some differences that mere repetition of arguments will not settle.

Regards

Stuart
--
Hewlett-Packard Limited registered Office: Cain Road, Bracknell, Berks
RG12 1HN
Registered No: 690597 England
Received on Wednesday, 24 October 2007 10:08:29 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:56:18 UTC