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

Re: Action: #rdfms-identity-anon-resources questions

From: Graham Klyne <Graham.Klyne@Baltimore.com>
Date: Thu, 26 Jul 2001 16:57:21 +0100
Message-Id: <5.1.0.14.2.20010726161731.03b946a0@joy.songbird.com>
To: RDF core WG <w3c-rdfcore-wg@w3.org>
At 07:06 PM 7/20/01 +0100, Graham Klyne wrote:
>Per today's teleconference, these are the questions that I feel need to be 
>answered to resolve issue #rdfms-identity-anon-resources [1].
>
>
>1. Does the current model and syntax spec sanction the use of an anonymous 
>resource to convey anything other than an assertion of the existence of a 
>resource with the properties given?  c.f. Frank Manola's comment in 
>[2].  See also Brian McBride's survey of relevant parts of M&S [3].

I think not.


>2. Does the use of an anonymous resource as a query variable represent an 
>extension to the idea an assertion of the existence of a resource with the 
>properties given?  Opinions vary: e.g. [4-9].

I think this is an extension to the idea of expressing existence.


>3 If the answer to Q1 is "no" and to Q2 is "yes", should we undertake to 
>extend M&S to allow this usage?  If so, why and how?

I think such extension would have far deeper ramifications than we can 
currently see;  e.g. maybe requiring introduction of universally quantified 
variables.  I am also thinking there is a possible relationship with 
use-without-assertion of RDF statements, which is currently a murky area.

My (tentative) answer, then, would be no, don't try and do this now (but 
note the issue for further exploration in RDF V2.0.


>4. Can we agree whether or not unique generated IDs (in the style of 
>Skolem constants) are equivalent to existentially quantified variables for 
>the purpose of asserting the existence of a resource with properties 
>given?  (See Frank's message [10] for a discussion -- I discount the 
>option of dropping anonymous resources.)

At the level of making simple assertions using the existential-conjunctive 
subset of FOL, I think they are equivalent.  But I'm not sure there won't 
be future developments of RDF that might expose subtle differences.


>If we agree they are equivalent:
>
>4.1 What are the advantages and difficulties with each approach?
>
>4.2 Which approach do we prefer?
>
>4.3 How does the preferred approach affect M&S?

I lean toward having a special "variable name" form along the lines of 
_:genid, for two reasons:
(a) it's easier, more compact and ultimately more dependable then trying to 
generate unique identifiers.  (Unique identifier algorithms are statistical 
rather than definite in their performance.  And the statistics aren't 
always as favourable as one might expect:  see, e.g., appendix A of RFC 2938.)
(b) it seems possible that differences will emerge if future developments 
extend the logical capabilities of RDF.

...

I think that the variable name form would (for now) behave like a URI in 
most respects, except one.

Assuming an operation that takes two independent N-triple expressions (= 
set/bag of statements) and combines them into a single N-triple expression 
containing all of the statements from each, names of the special form are 
changed/reassigned so that they are unique in the new expression;  this 
description needs tightening up to maybe something like:

(1) two variables are DISTINCT if:
     (a) they have different names, OR
     (b) they appear in different N-triples expressions

(2) when combining two N-triples expressions into a single expression, all 
variables are assigned arbitrary names (which may be the same as their 
original names) such that DISTINCT variables always have different names in 
the resulting expression.

By this approach, the variables are maintained with distinguishable forms, 
but may be treated equivalently for the purposes of the model 
theory.  Processing of such variables is a syntactic operation on N-triples 
expressions.


>5. Are there any other uses of anonymous resources that have been overlooked?

I see none that are consistent with my understanding of the expressive 
capability of RDF 1.0.

#g
--


>[1] http://www.w3.org/2000/03/rdf-tracking/#rdfms-identity-anon-resources
>
>[2] http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0124.html
>
>[3] http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0244.html
>
>
>Opinions on representation of query variables:
>
>[4] Dan 
>C:  http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jun/0435.html
>
>[5] Pat 
>H:  http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jun/0543.html
>
>[6] Pat 
>H:  http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jun/0518.html
>
>[7] Dan 
>C:  http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jun/0537.html
>
>[8] 
>Frank:  http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jun/0538.html
>
>[9] 
>Frank:  http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jun/0559.html
>
>
>[10] http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0218.html
>
>
>
>------------
>Graham Klyne
>(GK@ACM.ORG)

------------------------------------------------------------
Graham Klyne                    Baltimore Technologies
Strategic Research              Content Security Group
<Graham.Klyne@Baltimore.com>    <http://www.mimesweeper.com>
                                 <http://www.baltimore.com>
------------------------------------------------------------
Received on Thursday, 26 July 2001 11:58:53 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:38:17 EDT