Re: resources, stuffs and individuation

* Roy T. Fielding <fielding@apache.org> [2003-04-23 03:11-0700]
> >The technology will change if we restrict "resource" to only refer to 
> >things that are in some sense 'identified' by a URI. That has 
> >immediate technological consequences for a whole range of 
> >applications, or else we will have to rewrite several draft W3C specs.
> 
> Which is, of course, why the current definition does not do that.

I believe many members of the RDF developer community, doing their best
to take 2396 on faith, will be glad to hear this stated so explicitly. 

[...]
> >>If you can come up with
> >>a better definition and can get rough consensus that it doesn't 
> >>exclude
> >>things that others consider to be resources, then I'll put that in the
> >>specification.
> >
> >I would prefer to not exclude anything. What's wrong with. "A resource 
> >can be anything."  I can't see any reason so far to impose any kind of 
> >restriction on what can count as one.

+1
 
> Anything includes those things that cannot be identified, and those 
> things are not resources.

Are there such things? To claim so without adopting some specific 
theory of reference / naming is a little vague. If my time in the Philosophy 
dept taught me anything, it was that theories of reference are subtle 
and unrewarding and liable to cause controversy. Which theory of reference 
we adopt will influence our judgement about which things (if any) are
unidentifiable. I find little practical engineering value in 
the mailto:uri@w3.org list engaging in discussions about "what counts 
as being identifiable?", yet we're inevitably drawn to that by 
making a thing's 2396-resource-ness dependent on whether or not it 
can "be identified", whatever that means. From where I'm standing, the 
simplest option seems to be to retreat to plain old 'resources are things'.
(this from someone who fled academic philosophy to avoid spending time on
futile debates about meaning and reference (doh! ;)).

Could you expand a little on this notion of 'things that cannot be 
identified'? Specifically, on why there is value in having a technical 
term for 'everything except things that can't be identified'.
...by whom? ...when? etc too -- can a thing become a resource? can a 
resource slip from resource-hood and become unidentifiable? 

I'd be particularly interested in use cases for where such distinction
has proved valuable in a Web context, and (if this even makes sense) some 
examples of unidentifiable things? Could you describe a few of them?
 
> >>http://lists.w3.org/Archives/Public/www-tag/2002Jul/0128.html
> >
> >Thanks for the pointer, but Im afraid that only makes things worse for 
> >me. I have no idea what you are talking about in this message. What 
> >for example is a 'system of identification'? What is 'sameness of 
> >essential character'?  Does it differ from inessential character, for 
> >example? Citing Webster isnt good enough, sorry: we are trying to get 
> >these ideas tied down well enough so that we can use mathematics on 
> >them. (BTW, if this is not a philosophical discussion, I must be 
> >dreaming.)
> 
> Heh, that's amusing.  I'd cite the OED or some more definitive source,
> but they don't have a convenient website.
> 
> Mathematics is a tool, not an answer.  If you can't do the translation,
> then maybe you are using the wrong tool.  What good is having a 
> mathematical
> model if that model fails to correspond to what is supposed to be 
> modeled? Maybe the wrong type of mathematics is being applied.

I think this tool could be applied with a bit more enthusiasm first, before
we admit defeat and agree that 2396's notion of 'identity' is just too 
subtle and rich for the clumsy mechanisms of mathematical logic.

Re the 'wrong type of maths', yes there are aspects of the URI story
eg dealing with change over time which may mean we go shopping for a 
formalism designed for representing such things. That is why it is imho 
important to get answers to the questions Pat raised such as 'can something
stop being a 2396:resource?'. I'm not sure what the process is for doing so, 
except proposing textual changes to the spec. Formalising the temporal 
aspects would likely be a nightmare, which is one of the reasons I'm 
happy with just saying 'everything is a resource, the end.'

FWIW we went through the same process with RDF. The 1999 spec was interesting
and evocative and specified in prose. Implementor feedback led us to go back
and clarify things with a combination of logic and test cases, I think with 
some success. There was similarly 

 
> >What do you mean by an 'identification mechanism'? (mechanism??)
> 
> A system of identification that has been mechanized.

Turned into a machine or mechanical process? So that specific acts of 
identifying can be undertaken without sentient supervision? I'm not sure 
I follow, could you give some examples of such mechanisms?

> >Your example of the clock I find particularly confusing. Why would it 
> >be a different resource if it gives me UTC time rather than local 
> >time? The criteria for what counts as a resource seem to be getting 
> >murkier with every sentence in your explanation, rather than clearer. 
> >(BTW, can the time itself be a resource?)
> 
> What is time?  Please don't tell me it is an absolute quantity 
> independent
> of any frame of reference.
> 
> >Im not trying to be difficult, honestly. I am just totally confused. 
> >None of what you say about resources enables me to get a clear picture 
> >of what you mean by the term.
> 
> I think you need to slow down and spend more time thinking about it.
> It's not as if this is a new subject, and I don't see any reason why
> you need to completely understand it in an day's study.

I believe Pat has more than done his reading and research, both in terms 
of ploughing through relevant W3C and IETF specs, following mailing 
list discussions, and through spending many years studying the 
subtleties of building formal representational systems. He asked good, 
if slightly bemused sounding, questions that may help tease out some  
of the unarticulated design principles and assumptions behind 
the RFC 2396 document.

> >Let me ask: why do we need to attempt to define or limit the term? 
> >What is wrong with just saying, a resource can be anything?
> 
> Because a resource must be identifiable.

I've just re-read 
http://lists.w3.org/Archives/Public/www-tag/2002Jul/0128.html
and am still puzzled. If by identifiable you mean 
"in principle identifiable" rather than "in actual fact 
identifiable 'cos it has a known URI", then you really do 
raise expectations that those principles be articulated somewhere.

Maybe we could have a URI FAQ:

(this is proposed draft text for the revised spec btw):
[[
Appendix X: Some common questions about Resource-nature...

Q: Are all things 'resources'?
A: No, only some things. Unidentifiable things are not 'resources'.

Q: Which things cannot be identified?
A: Only those things which, in their current history, have 
   not been named or in-principle nameable with a URI reference.

Q: What is it to be name-able by a URI reference?
A: To be the kind of thing that can truly be said to be named
   by an RFC2396-formatted URI reference string.

Q: Are there any things that are not in this category?
A: At the present time, we do not know.

Q: Can an unidentifiable thing become identifiable?
A: Yes, either through improvements in Web naming 
    infrastructure, or through changes in its 
    characteristics.

Q: Are there any things which are currently 'resources'
   but which were previously not identifiable.
A: This is an open research question.

Q: Can a resource ever stop being a resource, yet 
   continue to exist as a non-resource thing.
A: Only in the event of a catastrophic collapse in 
   the in-principle identifiability infrastructure.
]]


etc. Ok the last answer was a bit flippant, but in general I'm 
serious. These questions come up all the time, and the spec 
rather than scattered mailing list archives seems a better 
home for their answers.

Dan

Dan

Received on Wednesday, 23 April 2003 15:37:37 UTC