W3C home > Mailing lists > Public > semantic-web@w3.org > March 2009

RE: first and rdf:rest as functional property

From: Jeremy Carroll <jeremy@topquadrant.com>
Date: Mon, 23 Mar 2009 09:40:15 -0700
To: "'Pat Hayes'" <phayes@ihmc.us>, 'Reto Bachmann-Gmür' <reto.bachmann@trialox.org>
Cc: "'Michael Schneider'" <schneid@fzi.de>, "'Bijan Parsia'" <bparsia@cs.man.ac.uk>, "'Story Henry'" <henry.story@bblfish.net>, "'Semantic Web'" <semantic-web@w3.org>
Message-ID: <000101c9abd6$0c86c640$259452c0$@com>

I tend to Reto's point of view that interoperability concerns lean towards a conservative use of lists.
(Although I feel very uncomfortable with that - trying to stick to the spirit of a rule that hasn't been well-formulated - it's soul-destroying - the secret policeman inside is so much more vicious than the one who is publicly accountable. In this case, we could end up with an incredibly minimalist view of a list, that obligates us, on their construction, to much more work than any recipient of rdf lists actually requires. Maybe RDF Core should have made stronger statements about lists: e.g.

[[
For an RDF Graph published in an RDF/XML document, the following SHOULD all hold:
+ For any blank node or uri node that is the subject of an rdf:type rdf:List triple, or the subject or object of an rdf:rest triple, or the subject of an rdf:first triple then
   - it is rdf:nil, and not the subject of an rdf:first or rdf:rest triple
  Or
   - it is the subject of precisely one rdf:first triple
   - it is the subject of precisely one rdf:rest triple 
+ there is no rdf:rest directed cycle 
]]
(such graphs are necessarily finite)
).


Also, I would suggest that lists, in practice, are syntactically functional - and that this isn't a semantic constraint and trying to express such a semantic constraint is an amusing indulgence, and not a practical utility. 
I would also reiterate that two lists sharing a tail is not, in my mind, a violation of conservative use of lists, but something that one should expect.

Jeremy




> -----Original Message-----
> From: Pat Hayes [mailto:phayes@ihmc.us]
> Sent: Monday, March 23, 2009 8:05 AM
> To: Reto Bachmann-Gmür
> Cc: Jeremy Carroll; 'Michael Schneider'; 'Bijan Parsia'; Story Henry;
> 'Semantic Web'
> Subject: Re: first and rdf:rest as functional property
> 
> 
> On Mar 23, 2009, at 6:43 AM, Reto Bachmann-Gmür wrote:
> 
> > Thank you all for your contributions to these thread. I'll try to
> > summarize how I understood things with regards to the original
> > questions:
> >
> >> My questions:
> >> - Are there useful usages where an rdf:list has several distinct
> >> rdf:first and rdf:rest value?
> >>
> > RDF Semantics speaks about "well-formed" list to refer to lists for
> > which the above (and possibly other) conditions are true. RDF however
> > does not impose "well-formedness" so it would be possible write
> > 'surprising', 'elegant', 'creative', 'unconventional', 'imaginative'
> > or
> > 'beautiful' lists that do not match the condition. However it
> > explicitly
> > mentions that extensions may exclude lists that violate the
> convention
> > so that creating beautiful non-regular lists is an interoperability
> > bug.
> >
> >
> >> - Is it just not written that rdf:first and rdf:rest are functional
> >> (maybe due to some spec layering reasons) or is false to consider
> >> rdf:first and rdf:next as functional?
> > There's currently no specification asserting that they are, but as
> RDF
> > semantics explicitly allows well-formedness limitations in semantic
> > extension there statements are not false (as this would contradict
> > the
> > semantic extensions).
> >
> >
> > To me this situation this situation is not really satisfactory.
> > Asserting:
> > |
> > _:666 rdf:first <ex:aaa> .
> > _:666 rdf:first <ex:bbb> .
> > _:666 rdf:rest rdf:nil .
> > <ex:aaa> ||owl:differentFrom <ex:bbb>
> >
> > |would be an interoperability bug as it would be a contradiction with
> > possible semantic extensions
> 
> It is impossible to guarantee compatibility with all semantic
> extensions, as they will be inconsistent with one another. Different
> versions of OWL are already inconsistent with one another. Indeed,
> this is one reason why the RDF specs take a minimalist approach to
> specifying anything at all controversial (such as axioms for lists.)
> 
> > while on the other hand we do not currently
> > have spec legitimation to draw the following conclusion:
> >
> > |_:666 rdf:first <ex:aaa> .
> > _:666 rdf:first <ex:bbb> .
> > _:666 rdf:rest rdf:nil .
> > =>
> > <ex:aaa> ||owl:sameAs <ex:bbb>|
> 
> But there is nothing to stop you declaring that you will make this
> inference, and making it. You should not seek to find spec
> legitimation: no specification can determine all your ontological
> decisions. Bear in mind that we are here talking about an RDF
> _ontology_ of "lists", not about list data structures themselves.
> 
> Pat Hayes
> 
> >
> > Even if asserting functionality of rdf:first and rdf:rest doens't
> rule
> > out all usage of the properties that result in non standard lists, it
> > would in my opinion be a benefit for the interoperability of
> different
> > system if these statements would be asserted by semantic web
> > standards.
> > Any chance that a semantic extension as foreseen by RDF Semantics
> > becomes part of the semantic web specifications?
> >
> > Cheers,
> > Reto
> >
> >
> 
> ------------------------------------------------------------
> IHMC                                     (850)434 8903 or (650)494 3973
> 40 South Alcaniz St.           (850)202 4416   office
> Pensacola                            (850)202 4440   fax
> FL 32502                              (850)291 0667   mobile
> phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
> 
> 
> 
> 
Received on Monday, 23 March 2009 16:41:12 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:28 GMT