SV: SV: SV: SV: FW: [plcs-dex] External_class_library usage

[<mats.nilsson@fmv.se>]

Even if all of the questions doesn't yet have definitive answers, I agree that a summary of this discussion would be a great idea. I think that the identification of a number of existing questions is a good result in it self, which we can continue to work with, now when it looks like we're about to start working within OASIS again.

Regards,
  Mats

-----Ursprungligt meddelande-----
Från: David Price [mailto:david.price@eurostep.com] 
Skickat: den 23 december 2006 03:27
Till: plcs-dex@lists.oasis-open.org
Ämne: Re: SV: SV: SV: FW: [plcs-dex] External_class_library usage

Hi Peter,

You can use multiple ontologies in the same exchange file. However, if you 
ever want to do any validation then you have to define one that imports all 
the others. So, your example (altered to use id not name):

External_class.id = urn:plcs:rdl:std#Serial_identification_code
External_class_library.id = urn:plcs:rdl:myproject:refdata.rdl

is exactly what I've been saying we should do provided that 
urn:plcs:rdl:myproject:refdata.rdl imports urn:plcs:rdl:std either directly 
or indirectly. This tells a post-processor that the classes that can be used 
in the exchange file are all defined in or imported into 
urn:plcs:rdl:myproject:refdata.rdl. You only need on instance of 
External_class_library per file in this approach.

Also, urn:plcs:rdl:myproject:refdata.rdl can import any number of ontologies 
and so 

External_class.id = http://www.dod.mil/dodaf/sv1/WeaponsSystem

can be used too, as long as http://www.dod.mil/dodaf/sv1 was imported  into 
urn:plcs:rdl:myproject:refdata.rdl as well.

These ontologies also have to specify the PLCS entity types to which the 
classes are applicable. That's why you can't just pick any old class you want 
and put its URI in for External_class.id ... you'd have no way to validate it 
was being used correctly.

Conceptually, this approach is a lot like if you implemented the short form of 
an EXPRESS schema. Because of the distributed nature of OWL, this approach 
works quite easily in that language where it's very difficult to deal with in 
EXPRESS (e.g. you have to import an entire ontology but you can use/reference 
subsets of an EXPRESS schema).

I think I need to write this up more carefully so it makes sense to everyone. 
It's hard to follow/explain in a string of emails.

Cheers,
David

On Friday 22 December 2006 19:27, Peter Bergström wrote:
> Thanks Dave, I'm happy I had a simple question for once...
> But...
>
> If the context is built up of a hierarchy of 'contexts' that inherits from
> each other, I can very well think of 'multiple context ontologies':
>
> Just for the sake of discussion (Christmas is coming...), consider a data
> set where I have a part classified as 'engine' by the OEM, and the
> Product_as_individual I'm using I have decided to classify as 'MainEngine'.
> The part and real object are two different objects, created at different
> points in time by different organizations, but I want the entire
> information overview so I have decided to include my OEM's classifications
> etc in my data set. Why can't I say that 'engine' has the context of
> 'OEMrdl', while 'MainEngine' has the context of 'MyOwnRdl'? If I could, I
> wouldn't have to import my OEM's rdl into my own rdl...
>
> You don't have to view this as an oxymoron, since these are two different
> objects. And my example here was poorly invented, I think there are far
> better examples if somebody else gave it some more thoughts...
>
> What I would have like to be able to do (as we do now) is to within *one*
> context (myproject) use identify those classes I use as defined somewhere
> else as belong to that 'somewhere else', even if they are in fact, due to
> import, part of the myproject context. This for the benefit of others to
> see and understand what may be common between my data and other people's
> data.
>
> I guess you wouldn't like this solution either:
>
> External_class.name = urn:plcs:rdl:std#Serial_identification_code
> External_class_library.id = urn:plcs:rdl:myproject:refdata.rdl
>
> Here I try to illustrate that the context ontology is myproject, but one
> specific object happens to be classified as defined in the oasis std
> ontology (which myproject is a specialization of).
>
> Am I cheating? Does the full path have to point at the only one context
> ontology as well?
>
> Peter
>
>
> -----Original Message-----
> From: David Price [mailto:david.price@eurostep.com]
> Sent: den 22 december 2006 18:27
> To: Peter Bergström
> Cc: plcs-dex@lists.oasis-open.org
> Subject: Re: SV: SV: SV: FW: [plcs-dex] External_class_library usage
>
> Hi Peter,
>
> At last a simple question...   For successful data exchange, all the rules
> for conformance, validation and interoperability have to be well-defined.
>
> This makes me think we're not using the term context in the same way. By
> context I mean the rules that enable interoperability. In the scenario you
> describe, you haven't specified sufficient information to guarantee
> validation of the exchange file. In fact, "multiple context ontologies",
> which to me is an oxymoron, is exactly the state of the currently specified
> External_class_library usage in DEXLib .... which I hope I've explained is
> insufficient.
>
> Cheers,
> David
>
> On Friday 22 December 2006 15:10, Peter Bergström wrote:
> > I'll look at the slides shortly, but I just have one single question
> > immediately:
> >
> > What is stopping me from having more than one context ontology? All
> > specialized terms would be referenced from the most specialized context
> > ontology, while the standard once would be referenced directly from the
> > standard context rdl.
> >
> > Saying:
> > > External_class_library http://www.dod.mil/dodaf/sv1/ and External_class
> > > WeaponsSystem
> >
> > Would then be unambiguous, pointing to the MoD definition,
> > while:
> >   External_class_library http://www.projecty.org and External_class
> >   WeaponsSystem
> > Would indicate a specialization (not necessarily, but you would have to
> > know the project y rdl to be sure).
> >
> > Question is however: Is it disallowed to use several context ontologies
> > in one data exchange, and if so, why?
> >
> >
> > Peter
> >
> >
> >
> > -----Original Message-----
> > From: David Price [mailto:david.price@eurostep.com]
> > Sent: den 22 december 2006 15:41
> > To: plcs-dex@lists.oasis-open.org
> > Subject: Re: SV: SV: SV: FW: [plcs-dex] External_class_library usage
> >
> > Hi Peter,
> >
> > I renamed the subject so make this easier to follow. I also deleted
> > everything except my example and your question.
> >
> > In the example from my previous email below, http://www.projecty.org and
> > http://www.projectx.org are the context ontologies. The context ontology
> > is the most specific, not the most generic.
> >
> > Attached are slides from 2004 presentations of the approach to the
> > Semantic Web and NASA/ESA PDE conferences. The MOD extension ontology is
> > the context ontology in this simple example. I added the
> > underlined/italics text today hoping to clarify the point.
> >
> > I guess I've been assuming these ideas had made it into the minds of the
> > PLCS RD developers and implementors ... clearly that was a bad assumption
> > on my part. Apologies about that.
> >
> > Cheers,
> > David
> >
> > On Friday 22 December 2006 13:07, Peter Bergström wrote:
> > > On the context ontology, can you give me some example of how this is
> > > supposed to be done, from a real world use of owl? Is the context
> > > ontology the most specific rdl, or the most generic (i.e. the PLCS
> > > express OWL schema)? Is the OWL import of all rdl's wrong approach, is
> > > that what you say? I may come out stupid here, but I can read your
> > > statements this way, or the totally opposite way... Excuse me, but I'm
> > > not a native English (American) speaking person... I probably need the
> > > picture book examples...
> >
> > No, the OWL is structured correctly wrt imports. See the attached slides.
> > The issue is the use of something to specify the context within which to
> > understand the RD classes that are used. I think External_class_library
> > makes sense for that purpose.
> >
> > > EXAMPLE
> > >
> > > Contractors cannot change the US DOD DODAF Ontology and there's a class
> > > there with URI http://www.dod.mil/dodaf/sv1/WeaponsSystem. In the
> > > current approach that's:
> > >
> > > External_class_library http://www.dod.mil/dodaf/sv1/ and External_class
> > > WeaponsSystem - but there's a problem lurking here.
> > >
> > > The statement that for Project X WeaponsSystem's a subclass of Product
> > > but for Project Y WeaponsSystem's a subclass of System_breakdown, but
> > > not Product, can't be defined in http://www.dod.mil/dodaf/sv1/ or
> > > anything it imports, it's specified in statements defined in the
> > > http://www.projectx.org and http://www.projecty.org ontologies which
> > > import it and PLCS. So the semantics of the use of
> > > http://www.dod.mil/dodaf/sv1/WeaponsSystem is unknown unless you know
> > > the exchange is in the context of the Project X or Project Y ontology.
> > > My point is that conceptually this is true for every class - even those
> > > specified in the standard PLCS RD because you don't know if subclasses
> > > of them have been defined so that the use of the superclass in the PLCS
> > > RD is not allowed. Specifying one instance of External_class_library
> > > that's the Project X ontology and using External_class.id as I've
> > > suggested solves this problem.
> >
> > This message contains information that may be privileged or confidential
> > and is the property of Eurostep Group. It is intended only for the person
> > to whom it is addressed. If you are not the intended recipient, you are
> > not authorized to read, print, retain, copy, disseminate, distribute, or
> > use this message or any part thereof. If you receive this message in
> > error, please notify the sender immediately and delete all copies of this
> > message.

-- 
Mobile +44 7788 561308
UK +44 2072217307
Skype +1 336 283 0606