RE: Yorick Wilks on Semantic Web & httpRange-14

[Copying to public W3C archive again.  Not sure if that's where you were
suggesting that we archive this.]

On Thu, 2012-05-17 at 08:47 -0700, Larry Masinter wrote:
> If it isn't clear: my thinking is that a URI means what it means,
> without any hope of ever having any other kind of "definition", and
> the notion that URIs can be used to mean something other than what
> they mean inside a@href is bound to failure.  "slithy toves" has a
> meaning only in a context, and there is no place in a well-architected
> web system for *another* channel of information besides the context of
> use.

I think there is a place, as the A, B, C use case in your last message
illustrates.  The other channel of information is the URI definition
provided by the URI owner.  That convention provides an efficient,
scalable way for parties A, B and C who know nothing about each other to
easily agree on a common definition if they choose to do so.  This is a
useful benefit, even if it does not go so far as to ensure that they are
all giving the same meaning to that URI.
> 
> If A says "slithy toves" to C and B uses the same term, and C wants
> further clarification of what A or B might have meant, the only
> authorities to ask are A and B. 

I agree.  That use case is way beyond what a convention like the Uri
Definition Discovery Protocol (UDDP)
http://www.w3.org/wiki/UriDefinitionDiscoveryProtocol 
attempts to address.

> There is no room for a (fourth) party, and what C will get from A and
> B are less "definitions" and more "clarifications": among all the
> things that "slithy toves" could possibly remind you of, which one did
> they (respectively) mean.

Yes, for that use case, there is no further need for the role of the URI
owner.  The URI owner only plays a role in the beginning, *before* the
parties have done any communication with each other.  If all parties use
the URI owner's URI definition, those parties are able to write their
RDF data in a way that makes that data easier to merge later.  The data
still can be inconsistent when merged, even if they all use the same URI
definitions, as illustrated in Figure 26:
http://dbooth.org/2010/ambiguity/paper.html#inconsistent-merge
But at least the range of potential inconsistencies has been constrained
by the use of common URI definitions.

I think you may be asking much more of a convention like the UDDP than
it was designed to address -- probably because it the problem has
wrongly (IMO) been framed in terms of "meaning".  It is true that this
convention does not nail down the meaning of a URI, nor does it help
communicating parties reach clarification when there is confusion or
misunderstanding.  It merely helps cooperating parties agree on common
URI definitions as a starting point.  But even this modest benefit has
proven to be useful in practice.  And these conventions would be even
more useful if they were were slightly better defined standardized.

David

> 
> 
> -----Original Message-----
> From: Larry Masinter 
> Sent: Thursday, May 17, 2012 8:38 AM
> To: 'David Booth'
> Cc: Henry Story
> Subject: RE: Yorick Wilks on Semantic Web & httpRange-14
> 
> You addressed my question, but your answer just raises more questions.
> Here's my question about URI definition discovery:
> 
> Party A and party B both communicate with Party C using the
> "semantic web". Party A uses the URI "slithy toves" when talking
> to C, and so does Party B. Now C wants to combine the data
> sets from A and B, and realizes that perhaps A and C mean
> different things by that URI. 
> 
> Now, somehow, you want some third party (oh, Lewis 
> Carroll, the "owner" of the URI "slithy toves") to tell you
>  the "definition" of "slithy toves"?
> 
> So that A and B can agree on definitions? So that C can
> discover what A and B meant?
> 
> Larry
> (mind if we archive this convo?)
> 
> -----Original Message-----
> From: David Booth [mailto:david@dbooth.org] 
> Sent: Thursday, May 17, 2012 6:50 AM
> To: Larry Masinter
> Cc: Henry Story
> Subject: RE: Yorick Wilks on Semantic Web & httpRange-14
> 
> Hi Larry,
> 
> On Wed, 2012-05-16 at 15:27 -0700, Larry Masinter wrote:
> > But I don't see how URI "definitions" have any role either.
> > A URI (as originally constituted) was a string that contained a recipe
> > for connecting to something on the internet, using ftp, telnet, email,
> > http, or any of a number of other connection methods.
> > So where do the "definitions" come from, and how do they fit in with
> > anything?
> 
> In the semantic web, we use URIs to denote things.  People want to be
> able to merge data such that when the same URI is used in multiple data
> sets, the same URI definition is also used for that URI, so that the
> data can be successfully merged.  This means that a convention is needed
> to enable independent parties to use the same URI definition.  That is
> the sole purpose of the idea of the "URI owner": it provides a way for
> independent parties to know which URI definition they should all use.
> None of this obligates anyone to use a particular URI definition, but
> for those who *wish* to use a common definition for a URI, a convention
> like this enables it.
> 
> Does that address your question, or did I miss what you were getting at?
> 
> 
> 

-- 
David Booth, Ph.D.
http://dbooth.org/

Opinions expressed herein are those of the author and do not necessarily
reflect those of his employer.

Received on Thursday, 17 May 2012 17:27:21 UTC