Re: [Moderator Action] Re: Discussing Valentina's and Aldo's work during AWWSW Telecon?

On Nov 26, 2008, at 12:15 PM, Harry Halpin wrote:

> On Wed, 26 Nov 2008, Jonathan Rees wrote:
>>
>> On Nov 24, 2008, at 10:36 PM, valentina presutti wrote:
>>> - we have put cnt:Content under ire:WebResource (which is what  
>>> Jonathan calls Accessible, i.e. AWWWIR)
>>
>> NO - these are completely different! See [1]. Yes, I think  
>> WebResource may coincide with Accessible = the things Pat says can  
>> be "accessed", which happens to coincide with "resource" in the  
>> HTTP/1.1 sense. I would say such things are concrete, ultimately  
>> physical, things, sort of like physical books - connected to space  
>> and time.
>
> I'm a bit confused about cnt:Content myself, but I'm also a bit  
> confused about Information Resource vs. Web Resource. A Web Resource  
> is an abstraction such a "web-page of the Eiffel Tower," and its  
> obvious that
> this should return a HTTP 200 in return to a HTTP GET.
>> But I defined "AWWWIR" to mean whatever AWWW means (which I've  
>> never been able to discern). Although I would be reluctant to say  
>> how these things relate to anything else, Tim, David, and I have  
>> now all agreed at least that whatever an AWWWIR is, it is very  
>> likely *not* an HTTP/1.1 resource = Accessible (this last equation  
>> is something we agreed made sense IIRC). Your "information object"  
>> seems to be not incompatible with an overlap with, or subsumption  
>> of/by, AWWWIR.
>
> Now, if Information Resources are disjoint with Web Resources and so  
> Web accessible things, then what are they? Can you  get a 200  
> response from an Information Resource? Yet, it's not "on" the Web?

Sorry, which definition of "Information Resource"? You are breaking  
the define-everything rule, and as moderator I insist you follow it.

> 	All we know about this category is  that if you get a HTTP 200 in  
> response from a GET on a URI, you are dealing with an information  
> resource.

There is not consensus in this group that httpRange-14 needs to be  
interpreted this way. You need to additional assumptions and a long  
chain of inference in order to conclude this. I'd rather stay away  
from this question and work on ontology instead.

> It seems to me like if an information resource is *both* something  
> like "Moby Dick the text" (which may not give a 200 since it's not  
> on the Web) or something like "Moby Dick the text on the web- 
> page" (which obviously should give a 200), then since "Moby Dick the  
> text on the web-page" is a Web resource, and since it returns a 200,  
> it seems the right answer is that Web Resources should be a subclass  
> of Information Resources.

If something can belong to both Accessible and Edition, then the  
classes overlap, which doesn't make sense. Perhaps there is a third  
class, of Accessible/Edition dumbbells, and that's what people mean  
when web-page URIs are used declaratively?

This is the elephant in the room.

I don't think it's up to us to say what others' URIs denote. Instead  
we should follow the AWWW URI ownership principle, that it is up to  
the naming authority to say what the URI denotes, as licensed by the  
applicable follow-your-nose URI scheme spec. Any intent that you  
impute to an HTTP response has to derive from something that the  
composer of the response would be willing to agree to. Normally this  
is HTTP/1.1 and a few other RFCs, period. If someone wants to say it's  
Accessible, fine. If they want to say Edition, well... we'd have to  
agree somehow that a "network data object" can be an Edition, which  
would be perverse. If they don't want to say, that's fine too.

If the URI owner says nothing, and then we start to say things that  
might impinge on what their URIs mean, then we must be very careful to  
respect their "URI owner" authority. (Not that I necessarily agree  
with "URI ownership" as specified, but it is part of the  
architecture.) We can say things such as "agent A accessed the  
referent of URI U thereby retrieving a representation" (AWWW-ese) that  
are inconsequential or empirical, but we shouldn't say anything a  
reasonable owner wouldn't be willing to agree to. Would they  
necessarily agree that their URI's referent is abstract, or that its  
essential characteristics can be conveyed? You would have to redefine  
these words as terms of art for them to do that. Then they will say  
"OK, say whatever you like in your private language, as long as it  
doesn't affect me."

I could go on and on. I believe we have a set of unsatisfiable  
constraints among HTTP/1.1, AWWW, authoritative metadata, follow-your- 
nose, httpRange-14, RDF semantics, and 14 billion URIs deployed in  
good faith. That's why we're fighting. Something has got to budge.  
Rather than go in circles we should come up with an ontology that,  
however ambiguous, can capture the basic features of ALL the specs and  
personal theories (like mine and David's). THEN we can see where the  
alignments and dissonances are among specs and between specs and  
practice, and which pieces of which specs need to be reinterpreted or  
extended. An OWL-DL reasoner might even help us out here by helping to  
track down the inconsistencies.

Jonathan

Received on Wednesday, 26 November 2008 18:55:21 UTC