Manditory vs. Optional graph names in TriX

> -----Original Message-----
> From: ext Jeremy Carroll [mailto:jjc@hplb.hpl.hp.com]
> Sent: 23 June, 2004 12:52
> To: Stickler Patrick (Nokia-TP-MSW/Tampere)
> Subject: Re: Extreme draft
> 
> 
> (may I resend this message to www-archive, Chris and Pat? 
> with a change 
> of subject line)
> 

Absolutely.

> 
> Your use case is fairly compelling ...
> 
> we could change the DTD and Schema back and add a note somewhere like:
> 
> [[
> TriX can be used as an alterative to RDF/XML
> for serializing unnamed graphs, simply by
> omitting the graph name. Such graph elements are
> understood in the same way as an RDF/XML document.
> ]]
> 

Exactly. This is how I have been thinking.

Names are new/optional, for new stuff like trust/provenance/quoting/etc.
but you can also benefit a great deal from TriX even if you don't care
about such new stuff.

> at some point we will end up having to address the underlying issue 
> though. The above para is a fudge, in my view, since it 
> doesn't tie in 
> with the abstract syntax or formal semantics.

I think it does tie in, in the sense that a graph with no name
is simply out of scope of the formal semantics.

I don't see how it conflicts with the abstract syntax since
any intersection of statements between an unnamed graph (which
cannot be described by RDF statements, even if it exists) and
a named graph is, as far as I can see, innocuous.

Names are optional in the TriX serialization, even if they would
be manditory for the semantics of named graphs (duh).

Eh?

Patrick

> 
> Jeremy
> 
> 
> Patrick.Stickler@nokia.com wrote:
> 
> > I.e., I should be able to have a command line tool that takes
> > an RDF/XML instance via stdin and converts the graph serialized
> > by that RDF/XML instance into a valid TriX instance.
> > 
> > If graph names are manditory, and there is no xml:base defined
> > for the root <rdf:RDF> element, I don't see how that is possible,
> > so that is not acceptable. (if you can see a way, other than
> > generating some ad-hoc name not represented in the input RDF/XML
> > itself, then great...)
> > 
> > If graph names are optional, then there's no problem.
> > 
> > Insofar as the semantics are concerned, a graph with no name is
> > simply outside the scope of RDF-based discourse, but still 
> serializable
> > with TriX.
> > 
> > Eh?
> > 
> > Patrick
> > 
> > 
> > 
> >>-----Original Message-----
> >>From: Stickler Patrick (Nokia-TP-MSW/Tampere) 
> >>Sent: 23 June, 2004 12:26
> >>To: 'ext Jeremy Carroll'
> >>Subject: RE: Extreme draft
> >>
> >>
> >>
> >>Hmmm... I'm still concerned about legacy RDF/XML documents not on 
> >>the web -- which do not necessarily have a name.
> >>
> >>I understand Pat and Chris' arguments about blank node names,
> >>but wonder if it is necessary to treat graphs with no names
> >>as having blank node names. 
> >>
> >>Graphs without names are just simply non-referencable in 
> RDF, period.
> >>But they should still be serializable using TriX.
> >>
> >>RDF is not intrisically tied to the web, so requiring backwards
> >>compatability between RDF/XML instances and TriX to only be in
> >>terms of web accessible URIs is, sorry, unacceptable to me.
> >>
> >>If we were not to posit a blank node name for graphs having no
> >>explicit URI name, would that be OK?
> >>
> >>???
> >>
> >>Patrick
> >>
> >>
> >>>-----Original Message-----
> >>>From: ext Jeremy Carroll [mailto:jjc@hplb.hpl.hp.com]
> >>>Sent: 23 June, 2004 12:03
> >>>To: Stickler Patrick (Nokia-TP-MSW/Tampere)
> >>>Subject: Re: Extreme draft
> >>>
> >>>
> >>>
> >>>Yes - I thought we had agreed it was now obligatory - 
> >>
> >>that's what the 
> >>
> >>>abstract syntax has; we modified all the examples to 
> >>
> >>include one. In 
> >>
> >>>terms of the migration path, any RDF/XML document on the web 
> >>>has a URL. 
> >>>I don't think there is much of a compatility issue. 
> >>
> >>Originally we had 
> >>
> >>>omitting a name as equivalent to having a new blank node 
> >>>name, but Pat 
> >>>and Chris shot those down.
> >>>
> >>>Jeremy
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>Patrick.Stickler@nokia.com wrote:
> >>>
> >>>
> >>>>Oops. I did miss one bug in the TriX DTD and Schema
> >>>>that is a change from the originally submitted version
> >>>>of the paper. 
> >>>>
> >>>>The latest version defines the graph name/URI as manditory, 
> >>>>but it should be optional in order to be compatible with the 
> >>>>existing RDF MT.
> >>>>
> >>>>I.e. the DTD says
> >>>>
> >>>><!ELEMENT graph        (uri+, triple*)>
> >>>>
> >>>>but should (and originally did) say
> >>>>
> >>>><!ELEMENT graph        (uri*, triple*)>
> >>>>
> >>>>And the Schema says
> >>>>
> >>>>  <element name="graph">
> >>>>    <complexType>
> >>>>      <sequence>
> >>>>        <element ref="trix:uri" minOccurs="1" 
> >>>
> >>>maxOccurs="unbounded"/>
> >>>
> >>>>        <element ref="trix:triple" minOccurs="0" 
> >>>
> >>>maxOccurs="unbounded"/>
> >>>
> >>>>      </sequence>
> >>>>    </complexType>
> >>>>  </element>
> >>>>
> >>>>but should (and originally did) say
> >>>>
> >>>>  <element name="graph">
> >>>>    <complexType>
> >>>>      <sequence>
> >>>>        <element ref="trix:uri" minOccurs="0" 
> >>>
> >>>maxOccurs="unbounded"/>
> >>>
> >>>>        <element ref="trix:triple" minOccurs="0" 
> >>>
> >>>maxOccurs="unbounded"/>
> >>>
> >>>>      </sequence>
> >>>>    </complexType>
> >>>>  </element>
> >>>>
> >>>>Note that the optionality if the name is what is stated
> >>>>in the prose content (in both the orignal and latest version):
> >>>>
> >>>>"TriX provides for graph naming by the use
> >>>>of an optional uri element before the triples
> >>>>of a graph."
> >>>>
> >>>>I know we had one or more discussions about this over the past few
> >>>>months, but thought we had agreed that the graph name was optional
> >>>>(even if an application might choose to attempt to derive a graph
> >>>>name in some way, such as using the URI via which the graph is
> >>>>obtained, or an xml:base value, etc.).
> >>>>
> >>>>Granted, cool stuff such as signed graphs and whatnot 
> >>
> >>cannot be done
> >>
> >>>>without naming the graphs -- but that is a requirement for 
> >>>
> >>>the application
> >>>
> >>>>of named graphs, and not of TriX -- and TriX must be fully 
> >>>
> >>>compatible
> >>>
> >>>>with the existing RDF specs. Making graph names manditory 
> >>
> >>IMO makes
> >>
> >>>>TriX incompatible as it does not allow the serialization 
> >>
> >>of existing
> >>
> >>>>graphs which reflects precisely the same information 
> >>
> >>which would be
> >>
> >>>>embodied in an RDF/XML serialization. If graph names are 
> >>>
> >>>manditory, then
> >>>
> >>>>a TriX serialization will always contain more information 
> >>>
> >>>than an RDF/XML
> >>>
> >>>>serialization of the same graph, since the former will have 
> >>>
> >>>a graph name
> >>>
> >>>>and the latter will not (at least not explicitely).
> >>>>
> >>>>Would you be opposed to reverting the DTD and Schema to 
> >>
> >>reflect the
> >>
> >>>>optionality of the graph name? 
> >>>>
> >>>>Patrick
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>>-----Original Message-----
> >>>>>From: ext Jeremy Carroll [mailto:jjc@hplb.hpl.hp.com]
> >>>>>Sent: 22 June, 2004 18:02
> >>>>>To: Stickler Patrick (Nokia-TP-MSW/Tampere)
> >>>>>Subject: Extreme draft
> >>>>>
> >>>>>
> >>>>>
> >>>>>Hi Patrick
> >>>>>
> >>>>>here is a draft, which I will send on Thursday if I have 
> >>
> >>not heard 
> >>
> >>>>>otherwsie.
> >>>>>
> >>>>>Changes:
> >>>>>
> >>>>>1) added "which is called TriX"
> >>>>>2) added forward ref from intro to syntactic extensions section
> >>>>>3) added para on accepted subset in formal semantics section
> >>>>>4) change a Dave Beckett ref
> >>>>>5) added <web> </web> markup around URLs in biblio
> >>>>>6) fixed DTD validity problems (quite a few)
> >>>>>
> >>>>>Jeremy
> >>>>>
> >>>>
> >>>>
> >>>
> > 
> 
>

Received on Wednesday, 23 June 2004 06:13:10 UTC