Re: Example for consideration: Resource versus Representation

Jonathan Rees wrote:

> Let's focus on trying to draw out Noah (and Tim) on the negative 
> assertion "a representation is not a Kwyjibo". Negative statements, 
> or conditions of potential inconsistency, are always much more 
> interesting and informative than positive ones. We just have to 
> figure out what the class Kwyjibo is supposed to be.  Noah, did you 
> mean one of the following, or yet a new concept?

I sort of put this thread aside while dealing with other things, and now 
notice that Jonathan is specifically asking my opinion.  I'm glad to try 
to give one, but doing so is complicated a bit by my having no clue what a 
Kwyjibo is.  Also, for better or worse, my background is much more in the 
direction of computing systems than formal semantics, which may tempt me 
to be a bit more mechanistic in my approach to things than you all will 
find helpful.  Anyway, the good news is that Tim's reply at [1] pretty 
much sums up what I would have said.  I should probably leave it at that, 
but since you asked me in particular, I'll say a bit more.

First of all, there's a lot of discussion about what is and isn't a 
resource in general.  A very important discussion no doubt, but you're 
asking me about whether representations are resources, and I don't think 
my answer depends greatly on which of the answers you choose from the 
short list of "what's a resource in general?" positions.  The one 
exception would be that if you take the position that resources are only 
things for which a URI has actually been assigned, then we have to 
acknoweldge that in practice URIs are at best rarely assigned to 
representations.

Tim Berners-Lee wrote:

> The Representation is the pair of the response headers and the bits.
> A fleeting thing sent over the wire.

Yes, exactly. 

Does such a representation have a URI?  Not in general, but there are 
situations in which it might be useful to assign one.  I think I gave this 
example on the phone:  Let's say I'm building a system to monitor or debug 
 Web traffic.  I want to make RDF statements like "that system over there 
was slow yesterday."  Well, one of the things I might want to do would be 
to make an RDF statement like "That representation, the one that was sent 
to Tim at 3 PM yesterday, was the longest representation that anyone 
retrieved all day."  To make that statement, I should would assign a URI 
to the representation.  It's a resource.  I would then reference that URI 
in RDF statements.

Is the representation an Information Resource?  Hmm.  It's certainly 
close, in that as Tim says the representation is mostly response headers 
and bits that clearly can be encoded as computer messages, because indeed, 
the whole point of a representation is to be on the wire.  Then again, 
those bits don't necessarily capture the "essence" of the representation 
as having been fleeting in time, sent between two particular endpoints 
(perhaps through proxies), etc.  So, I'm a bit on the fence as to whether 
it would be OK to return a 200 on a GET to the URI assigned to the 
representation itself.   Certainly it's OK to deploy a server, perhaps as 
part of that Web monitoring system, that would at very least return a 303, 
redirecting to useful information in RDF about the representation (it was 
sent yesterday at 3PM, it was sent to Tim, it was very long, etc.)

So, I do think representations are (or can be) resources.  You can assign 
URIs to it, when you have reason to do so, and you can find use cases such 
as the one above in which that's a very helpful thing to do.  You can make 
semantic web statements about them by referencing the URI.

Noah


[1] http://lists.w3.org/Archives/Public/public-awwsw/2008Jan/0031.html

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------

Received on Friday, 15 February 2008 23:28:44 UTC