Re: Comments on "SPARQL 1.1 Uniform HTTP Protocol for Managing RDF Graphs"

moving to awwsw to continue, reply below

David Booth wrote:
> On Fri, 2011-03-18 at 20:51 +0000, Nathan wrote:
>> Here's another quick rewrite:
>>
>> David Booth wrote:
>>> On Fri, 2011-03-18 at 18:55 +0000, Nathan wrote:
>>>>> Can something be both a birth certificate and a red lightbulb? (my intuition says no).
>>> In a given graph g, a URI u can perfectly well (ambiguously) identify
>>> something that is both a birth certificate and a red
>>> lightbulb, provided that g has no disjointness or other such predicates
>>> that would prevent it from being so.
>>>
>>> You need to know what graph you are asking about, and what assertions it
>>> contains, to answer the question.
>> So perhaps the question being answered is, can we feasibly carry out a 
>> conversation where we refer to both a birth certificate and a red 
>> lightbulb by a single ambiguous name? using RDF?
> 
> Yes, if the distinction between birth certificates and red lightbulbs is
> irrelevant to our task.
> 
>> Possibly, but why even try?
> 
> Because: 
> 
> 1. Not all applications *need* to distinguish between a lightbulb and a
> birth certificate.  For example, an application that cares only about
> who owns what may work perfectly fine treating <http://example/#item437>
> as an ambiguous combination of a birth certificate and a red lightbulb.

Not sure I can agree with that, "who owns what?", and "what" is? the 
birth certificate or the lightbulb?

> 2. Because this kind of ambiguity of reference is *inescapable* (though
> the example is an extreme case), so we have no choice but to learn to
> deal with it.  

Is "ambiguity of reference" not itself ambiguous? do you mean ambiguity 
in the description of what is being referred to (not enough information 
to get a perfect description, as is always the case)?, or, the ambiguity 
of the name (the name is used to refer to two or more distinct things)? 
or, do you mean that two different names are used to refer to a single 
thing, and that's where the ambiguity is?

> Ambiguity/unambiguity is *relative* to a particular application.  It is
> not an absolute notion.  A resource definition that is unambiguous to
> one application may be ambiguous to another application that requires
> finer distinctions.  In all but vanishingly few cases, it is
> *impossible* to define a resource in a universally unambiguous way,
> because there are *always* finer distinctions that can be made.  

The wording of this reply is very interesting, because it feels to me 
that you are swapping between discussing two different things. Unique 
names and unambiguous descriptions, the latter of course can always be 
refined almost infinitely, and what is being referred to is open to 
interpretation, it may be conjure up the thought of one thing to one 
person and another thing to another person, you may learn more about a 
thing described over time, and combine information from different 
sources together to get a clearer picture and so forth.

Is this not a different issue though? Both human and machine 
descriptions of things are inherently ambiguous and open to ever finer 
distinctions as you say. Human names are also very ambiguous, when I say 
"David" to who do I refer, in which context? However, AFAICT, machine 
names, URIs, are pretty unambiguous and the web can be seen as a global 
conversation in which those names are used; sure there may be many 
different names used to refer to a single thing, but each name 
(regardless of context) is usually only (or can only) be used to refer 
to a single distinct thing.

That is to say, if each URI really does refer to multiple different 
things, then we can expect the web to fall apart pretty quickly, or not 
to work at all; for instance we'd expect this email to get sent to maybe 
anybody, maybe you, I'd send you a link and it would just go to any page 
on the web - I'm quite sure it would all fall apart and fully expect 
there to be no web if this were the case.

So how, or why, is RDF any different, why would anybody want to 
knowingly use a single URI to refer to two or more different things in 
machine readable data? Who is going to go out there and say:
  { :x a :LightBulb, :BirthCertificate }
and who is going to listen to them if they do?

As time progresses, I increasingly wonder how things have got to this 
stage, URLs refer to web pages and stuffs on the web, not monkeys, you 
can try and say your URL refers to a monkey but everybody else will be 
using it to refer to something else, namely the thing on the web about 
that monkey. Likewise one URL refers to one thing.

In all honesty though, the thing I really don't get the most, is why? 
for who's benefit are we doing this? which new-comer to the semantic web 
and linked data is going to be confused if we say "yeah, we use URLs to 
talk about things on the web like web pages and rdf documents, just like 
everybody else on the web"? Why all the soft-lining on this and why are 
we making up special rules that only certain people in specific contexts 
with special machinery will understand?

Sorry, mini-ranted there!

Best,

Nathan

Received on Saturday, 19 March 2011 02:41:07 UTC