[oslc-core] [rant] Re: OSLC Core spec query shapes example

Arthur Ryman ryman at ca.ibm.com
Wed May 12 08:45:29 EDT 2010 

Olivier,

You wrote: 
> In particular, I don't quite get it why one needs to reinvent the wheel
> with shapes in OSLC when there's already ontologies as OWL to describe
> resources on the Semantic Web... for instance.

I agree that we are reinventing wheels here. But we're not reinventing the 
ontology wheel. We are reinventing the XML Schema wheel. Let me explain.

We are defining REST APIs, i,e, these are programmatic interfaces, and 
whenever you define a programmatic interface it is useful to also provide 
a machine-readable description of the interface in order to support tool 
automation. For example, we'd like to be able to help users create queries 
so we need to know what properties are available.

In the previous generation, we used XML technologies like XML Schema and 
WSDL. However, these are not tuned to the current Linked Data movement 
which is based on REST and RDF. Moving to RDF naturally leads you to think 
that OWL is the logical equivalent of XML Schema. However, ontology 
languages like RDFS and OWL have a very different purpose than XML Schema.

XML Schema describes constraints on the shape of an XML document. You can 
validate a schema with respect to a schema. The schema is therefore a set 
of constraints that the document must satisfy.

In contrast, RDFS and OWL let you express inference rules. Given a set of 
triples and an ontology, you can infer more triples. For example, if my 
ontology says that the range of ex:teamMember is a foaf:Person, and I am 
given a triple like: 

org:oslcCoreWorkgroup ex:teamMember mailto:olivier.berger at it-sudparis.eu .

then I can infer that:

mailto:olivier.berger at it-sudparis.eu rdf:type foaf:Person .

The ontology does not in general constrain the triples the way that XML 
Schema constrains the shape of a document. We therefore needed the 
equivalent of XML Schema to describe the actual shape of resources.

OWL is fairly powerful and it has inference rules that can lead to 
detectable inconsistencies. For example, if you assert that two classes 
are disjoint but also assert that some individual is a member of both, 
then a reasoner could detect that inconsistency. It might be possible to 
express the semantics of our Shape resources in OWL by exploiting this 
aspect of OWL, but in the spirit of KISS we have stayed away from baking 
RDFS or OWL into the OSLC core.

Regards, 
___________________________________________________________________________ 

Arthur Ryman, PhD, DE


Chief Architect, Project and Portfolio Management

IBM Software, Rational

Markham, ON, Canada | Office: 905-413-3077, Cell: 416-939-5063
Twitter | Facebook | YouTube





oslc-core-bounces at open-services.net wrote on 05/12/2010 05:18:25 AM:

> [image removed] 
> 
> [oslc-core] [rant] Re: OSLC Core spec query shapes example
> 
> Olivier Berger 
> 
> to:
> 
> Dave
> 
> 05/12/2010 05:18 AM
> 
> Sent by:
> 
> oslc-core-bounces at open-services.net
> 
> Cc:
> 
> oslc-core
> 
> Hi.
> 
> Pardon me in advance for that "rant"... It's just meant to help clarify
> the current OSLC direction. Apologies in advance for the good people
> investing efforts in OSLC... I'm just a devil's advocate.
> 
> 
> Having had a look at the page below, I tend to think that it misses the
> rationale for all this.
> 
> And I'm a bit fearful that OSLC turns to a too complex thing to ever
> achieve some potential of being a standard in the ALM field, if it
> becomes too hard to grasp (at least for the core mandatory concepts).
> 
> 
> In particular, I don't quite get it why one needs to reinvent the wheel
> with shapes in OSLC when there's already ontologies as OWL to describe
> resources on the Semantic Web... for instance.
> 
> 
> At least for the first half of the document references below, it isn't
> clear. Now, when the format of the query results is described, it may
> make more sense... but so hard to grasp at first sight !
> 
> 
> I'm a bit afraid of too much over-specification for these
> shapes/queries... whereas it isn't even clear if there's a single
> consumer available somewhere (OK, I mean in the Open Source landscape)
> that supports at least OSLC-CM V1 at the moment.
> 
> 
> Ensuring OSLC becomes a standard (an Open one, I mean), means it may not
> need to go too far and too fast, in particular in what concerns
> over-specification for the core. I strongly believe bottom-up approaches
> are interesting, and not so sure this is happening here now.
> 
> 
> Pardon me if I haven't fully understood all that's happening (unability
> to participate in conf calls certainly doesn't help much, and focusing
> on coding (OAuth ATM) prevents me from reading all the docs in time),
> but I fear there's some confusion between requirements for one
> implementor's requirements design/specs, and specification of an open
> standard, that should be at least a little bit understandable by the man
> in the street.
> 
> I'm doubtful it will be really easy to convince people to implement so
> complex specs in existing tools, given the level of complexity I can
> feel around here.
> 
> Maybe a more KISS approach would be better until more feedback comes
> from the grassroots implementors ?
> 
> Hope this helps anyway ;)
> 
> 
> Best regards,
> 
> Le mardi 04 mai 2010 à 17:20 -0400, Dave a écrit :
> > To better understand how queries work in the OSLC Core spec today,
> > here's an example that shows how to define a blogging service with
> > blog entries, blog comments and a query capability that can query
> > across both types of resources.
> > 
> >   http://open-services.net/bin/view/Main/OslcCoreQueryDiscussion
> > 
> > Examples are provided in RDF/XML of these resources:
> > - Blog Entry, Blog Comment and Query Resource shapes
> > - Service Provider
> > - Query response
> > 
> > Thanks to Arthur for a quick review and the one query URI example in
> > the document.
> > 
> > Feedback is welcome, as are additional query URI examples.
> > 
> > Thanks,
> > Dave
> > 
> > _______________________________________________
> > Oslc-Core mailing list
> > Oslc-Core at open-services.net
> > http://open-services.net/mailman/listinfo/oslc-core_open-services.net
> 
> -- 
> Olivier BERGER <olivier.berger at it-sudparis.eu>
> http://www-public.it-sudparis.eu/~berger_o/ - OpenPGP-Id: 2048R/5819D7E8
> Ingénieur Recherche - Dept INF
> Institut TELECOM, SudParis (http://www.it-sudparis.eu/), Evry (France)
> 
> 
> _______________________________________________
> Oslc-Core mailing list
> Oslc-Core at open-services.net
> http://open-services.net/mailman/listinfo/oslc-core_open-services.net

More information about the Oslc-Core mailing list