Re: JAR's exploration of TimBL's notion of information resource

On Tue, May 26, 2009 at 3:44 PM, Jonathan Rees <jar@creativecommons.org> wrote:
> On Tue, May 26, 2009 at 6:14 AM, Harry Halpin <hhalpin@ibiblio.org> wrote:
>> I'm also going to make a quick note:
>>
>> I think a good deal of confusion could be avoided by just ditching the
>> idea of conceptualizing a resource as a set of representations,
>> perhaps indexed by time/HTTP requests. This is the "FTTR" hypothesis
>> (I'm not sure if Roy would actually agree this is his hypothesis
>> also). It's great what work JAR has done thinking through generic
>> resources and their parameters of possible wa:representations.
>> However, trying to define a resource as just a set of
>> wa:representations is not a really good idea without some more steps,
>> steps that may undermine "defining" the URI.
>
> The only person trying to do this was David. Tim has clearly said that
> the wa-representations do not in themselves determine the resource.
> I'm trying to draw
> a clear distinction by introducing the notion of a generic resource's "trace".
> The trace isn't the GR, any more than the graph of measurements of
> a room's humidity is the room's humidity.

That I agree with. The concept of a "trace" of a resource's
wa:representations is interesting and probably useful.

>> Here's a simple example - "relative URIs in wa:representations".
>>
>> You can have two *identical* sets of representations given in response
>> to a single URI. However, the URIs they link to, if they are using
>> relative URIs, are not identical and can change the meaning of the
>> document, as well as what you can access mechanically from it.
>>
>> Thus, to conceptualize representations properly, you would at least
>> have to absolutize all relative URIs.

I was just noting that this simple example helps clarify why resources
can't be sets of representations ala "FTTR" in a simple-minded
fashion. The relative URI issue at least complicates that, although
it's unclear if it complicates it enough to "kill" the idea.


> This seems like a good thing to model. But there has been general (and
> hard-won) agreement in this group that an HTTP entity is a
> wa-representation. You'd have to invent a new class whose individuals
> possess a binding for the base URI. We could call it an
> hh-representation...  (What is Moby Dick's base URI, though?)
>

Trust me, after having spent quite a while with base URIs in GRDDL,
all I can say is that it's a bit of a mess. Somewhere there is a
solution (we have it in GRDDL), but I remember it being very complex.
Also, for resources in the large, of course there aren't base URIs.
Thus, the interesting thought resources with URIs are a subset of all
resources.

Received on Tuesday, 26 May 2009 12:50:38 UTC