W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > July 2001

Re: A use case for anon nodes - action from telecon

From: Frank Manola <fmanola@mitre.org>
Date: Tue, 17 Jul 2001 16:40:34 -0400
Message-ID: <3B54A2C2.30609@mitre.org>
To: Brian McBride <bwm@hplb.hpl.hp.com>
CC: rdf core <w3c-rdfcore-wg@w3.org>
Brian--

I wonder if you could resend your example, annotating it with which 
things are supposed to be the anonymous resources, and which aren't?  I 
frankly (how else can I talk?) found it difficult to properly interpret 
it.  Some of the comments below may be based on the resulting 
misinterpretation.

Brian McBride wrote:

> Aaron,
> 
> The thing to bear in mind about this submission is that these
> were not my words.  They were the words (with some HP specific
> stuff removed) of my colleagues who are using RDF in the manner
> described.  This input is from real developers.
> 
> Aaron Swartz wrote:
> 
>> On Monday, July 16, 2001, at 02:23  PM, Brian McBride wrote:
>> 
>> 
>>> (1) In the seller advert it would appear that the seller is
>>> only advertising a
>>> single specific (but under-specified) service, #anon12345 or
>>> whatever, which
>>> would be hard to distinguish from an actual service instance
>>> like #service42.
>> 
>> Why would you want to distinguish between the two?
> 
> 
> I think the idea here is that there will be URI's
> denoting specific services.  My colleagues are 
> interpretting a node with a URI to be denoting 
> such a service.  It would be wrong to match a different
> service.
> 
> When an anonymous node is specified, then no such constraint
> exists.  Thus a processor would process these two instances
> differently.
> 
> 
>> And I see
>> nothing about a URI that licenses you to assume that there is
>> only such thing.
> 
> 
> Oh we really do need this model theory don't we.  I tend to think
> of a URI as identifying one thing, and one thing only, but that way
> lies a philosphical debate on the nature of 'one'.  Shudder!

We certainly need the model theory, but while the model theory will help 
us clarify what we mean, it won't do all the work for us.

For example, there seems to be this assumption that an anonymous 
resource represents a variable in a query or template that is to be 
bound to specific instances (e.g., in the book seller example). 
However, I don't think this is the proper interpretation of any of the 
uses of anonymous resources in the M&S, the first one being "The 
individual whose name is Ora Lassila, email <lassila@w3.org>, is the 
creator of http://www.w3.org/Home/Lassila."  Rather, I think the basic 
interpretation of RDF statements is that of being assertions, and in 
some cases you wind up asserting the existence of something that you 
don't have a URI for, so you generate an identifier for it (e.g., the 
"individual" in the previous statement).  I amplified on this in one of 
my earlier messages [and I'm using the present message as a stake in the 
ground;  I'm composing a longer dissertation on this that I'll send 
later, since I have an action item on anonymous resources].

Your comment above about a URI only representing one thing suggests 
looking at the problem from the opposite direction too.  That is, while 
sometimes you may want to use an anonymous resource to refer to a "real 
thing" (like the "individual" whose name is Ora Lassila), in other cases 
you may want to use a genuine URI to refer to a template, or one of its 
parameters.  These are, after all, real things that you want to be able 
to talk about (particularly in the context of Web applications).  The 
fact is that there isn't really any built-in interpretation (at least 
not that I can see) that URIs only identify things that "really exist" 
(in some sense) and anonymous resources only identify variables to be 
matched.  Rather, these are interpretations we assign to them in 
specific uses.  I can easily imagine a "template" (e.g., a specific 
purchase order form) having a URI, with each of its component items also 
having a URI (I might want to describe its format and semantics, for 
example).  A specific instance of that form, partially or totally filled 
in with specific values, might have a separate URI.  There would 
probably be machinery for relating the template to the filled in 
instance, but it would be part of some application semantics, not 
necessarily built into the semantics of RDF.

 
>>> (2) Similarly in the buyer advert instead of describing a
>>> template, giving the
>>> service a URI would make it appear that I am looking for a
>>> specific service with
>>> that URI.
>> 
>> Umm, aren't you? What's the difference? In both you're looking
>> for something with these properties.
> 
> 
> The difference is that when a URI is specified the assumption is
> that someone, somewhere has defined a 'well known' name for this
> service.
> 
> The essence of this issue seems to involve the idea that the
> act of naming something in the internet is somehow, special.
> That if a processor is told that something has URI ISBN-12345
> or whatever, it had better not match that with anything that
> it does not 'know' is named ISBN-12345.  On the other hand,
> if a node is not named, then it can be matched with anything
> that matches its properties.
> 
> I'm not sure that standard FOL captures this.  FOL is built
> around a conceptual model where there can be many interpretations
> for statements in the FOL.  But that is not the situation we
> are in here.  We have one interpretation - its a mapping to 
> the world out there.  Intuitively I feel that makes a difference.
> That makes naming special.
> 

I'm not sure what you mean here.  If you mean that FOL statements can 
refer to both real things and "conceptual" things like templates or 
queries, that's true (e.g., I can describe the contents of a query in 
FOL), but, as I tried to argue above, so can URIs.  What seems to be 
happening in these examples, though, is that we're taking some syntax, 
and interpreting it in slightly different ways depending on whether we 
think we're talking about a query, or about a set of assertions.  I 
discussed this in 
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jun/0559.html
I think we need to be very careful about this.

This isn't to say that there might not be a place for variables in RDF. 
  I just don't think that that's automatically the appropriate 
interpretation of anonymous resources.

--Frank



-- 
Frank Manola                   The MITRE Corporation
202 Burlington Road, MS A345   Bedford, MA 01730-1420
mailto:fmanola@mitre.org       voice: 781-271-8147   FAX: 781-271-875
Received on Tuesday, 17 July 2001 16:43:09 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:38:11 EDT