W3C home > Mailing lists > Public > public-sdw-wg@w3.org > November 2016

RE: rdfs:class versus owl:class in SOSA-Core & JSON

From: Kerry Taylor <kerry.taylor@anu.edu.au>
Date: Sat, 12 Nov 2016 05:20:53 +0000
To: "janowicz@ucsb.edu" <janowicz@ucsb.edu>, Rob Atkinson <rob@metalinkage.com.au>, "Simon.Cox@csiro.au" <Simon.Cox@csiro.au>, "antoine.zimmermann@emse.fr" <antoine.zimmermann@emse.fr>, Armin Haller <armin.haller@anu.edu.au>, "jano@geog.ucsb.edu" <jano@geog.ucsb.edu>, "jlieberman@tumblingwalls.com" <jlieberman@tumblingwalls.com>
CC: "public-sdw-wg@w3.org" <public-sdw-wg@w3.org>
Message-ID: <PS1PR06MB174089204B968C0F83091A5FA4BA0@PS1PR06MB1740.apcprd06.prod.outlook.com>
I think Rob’s (2) below is the only thing that makes any sense, especially in the context of the decision at the F2f in Lisbon two have exactly 2 modules – core and ssn-minus-dul. (with the dul alignment also available, but not part of it).

I rather like Rob’s(1)  but it is much too late to get there from here. I can live with Rob’s(3).

But I am not sure I understand this:

Ø  Note also that if we publish two modules: sosa.OWL-DL and  sosa.RDFS,   and the RDFS is generated from the OWL, then the fact that RDFS does not have an import makes option that nearly equivalent to option #2 (the OWL does not need to import the RDFS, although it could do so)


From: Krzysztof Janowicz [mailto:janowicz@ucsb.edu]
Sent: Thursday, 10 November 2016 3:34 PM
To: Rob Atkinson <rob@metalinkage.com.au>; Kerry Taylor <kerry.taylor@anu.edu.au>; Simon.Cox@csiro.au; antoine.zimmermann@emse.fr; Armin Haller <armin.haller@anu.edu.au>; jano@geog.ucsb.edu; jlieberman@tumblingwalls.com
Cc: public-sdw-wg@w3.org
Subject: Re: rdfs:class versus owl:class in SOSA-Core

Hi Rob,

I think this goes back to the issue I raised here: https://lists.w3.org/Archives/Public/public-sdw-wg/2016Nov/0065.html . Lets discuss our modeling choices first and then decide about representational choices. The solution we have now is just fine and so far nobody was able to bring up any other argument that would point to any problem (assuming we add both owl:class and rdfs:class as extra level of safety).

SOSA -core is supposed to be a conceptually simple, not based on a 'simple' representation language.


On 11/09/2016 08:03 PM, Rob Atkinson wrote:

my opinion is that I'm happy to have owl constructs as long as the rdfs that would be entailed by OWL reasoning, but not RDFS reasoning, is materialised, and hence the user is not expected to use an OWL reasoner.  There seems to be a consensus (or at least no counter-arguments to this specific policy ACAICT)

I am therefore agnostic as whether all OWL is in a extension model which supports OWL-DL (?) reasoning - and is part of the normative semantics. I guess the question is whether there is a need for a subset of OWL semantics which imposes a low burden an OWL reasoner (a simple core). Others with more experience need to weigh in and document the subset and we can then vote if we are happy to include it in the core.

Kerry has made the useful point that expectations are raised when a user finds OWL constructs. Perhaps that can be managed by clear enough documentation that all RDFS semantics implied by such OWL is materialised.

There is a possible lower level of commitment - in that in the core terms are just defined using SKOS, without any form of class model - but I think we all feel that the world is used to simple RDFS models and we could use that as a moderately expressive baseline.

So I see the choice is between:
1) something like SKOS Concepts to "reserve" resource names and attach documentation
2) RDFS only + OWL module that imports it
3) RDFS inc. "cheap to process OWL"  + OWL-DL module that imports it
4) OWL-DL only
5) OWL-DL with entailed RDFS
6) something I've missed :-)

Note also that if we publish two modules:
sosa.OWL-DL and  sosa.RDFS,   and the RDFS is generated from the OWL, then the fact that RDFS does not have an import makes option that nearly equivalent to option #2 (the OWL does not need to import the RDFS, although it could do so)

I propose we get this list of options nailed down so we can identify which flavour we are talking about in future, and then if we cant get a consensus quickly document pros and cons, and specifically anything we might break for users. (As an example I personally think #4) (OWL-DL only) breaks it for RDFS-only based reasoning)

Rob Atkinson

On Thu, 10 Nov 2016 at 12:55 Kerry Taylor <kerry.taylor@anu.edu.au<mailto:kerry.taylor@anu.edu.au>> wrote:
And therein lies another dragon -- or   an elephant?

In ssn  a ssn:MeasurementCapability   is not defined with domain/range. Instead it is defined via a local allvaluesfrom restriction on Sensor.  So certainly this following  holds:

>   ?x :hasMeasurementCapability  ?y .
>  I cannot conclude that ?y is a :MeasurementCapability

One of the many promises of the "simple"  core is that it is simpler than full  ssn in the sense that   it makes weaker  ontological commitments (see the notion of "vertical modularisation" in the FPWD).

So this automatically forbids rdfs:domain and rdfs:range in this case and similarly most others. Or otherwise it forbids one of many other broadly agreed goals.

Having said that, which remains true in general, hasMeasurementCapability  in particular is not currently  proposed to exist at all in sosa-core  https://github.com/w3c/sdw/blob/gh-pages/ssn/rdf/sosa.ttl.  So there may be no issue there.

However the same principle should indeed apply to observedProperty, for example, which does occur in the core.

-----Original Message-----
From: Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au> [mailto:Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au>]
Sent: Thursday, 10 November 2016 12:19 PM
To: antoine.zimmermann@emse.fr<mailto:antoine.zimmermann@emse.fr>; janowicz@ucsb.edu<mailto:janowicz@ucsb.edu>; Armin Haller <armin.haller@anu.edu.au<mailto:armin.haller@anu.edu.au>>; jano@geog.ucsb.edu<mailto:jano@geog.ucsb.edu>; jlieberman@tumblingwalls.com<mailto:jlieberman@tumblingwalls.com>; Kerry Taylor <kerry.taylor@anu.edu.au<mailto:kerry.taylor@anu.edu.au>>
Cc: public-sdw-wg@w3.org<mailto:public-sdw-wg@w3.org>
Subject: RE: rdfs:class versus owl:class in SOSA-Core

> something really wrong in the name of the property!  Or, alternatively there ought to be a rdfs:range axiom.

An owl:ObjectProperty without a rdfs:range axiom is usually be missing the most significant part of the semantics. I generally add these as a matter of course (not so for rdfs:domain). As Jano pointed out, semantics shouldn't depend on labels.


-----Original Message-----
From: Antoine Zimmermann [mailto:antoine.zimmermann@emse.fr<mailto:antoine.zimmermann@emse.fr>]
Sent: Thursday, 10 November 2016 10:06 AM
To: janowicz@ucsb.edu<mailto:janowicz@ucsb.edu>; Armin Haller <armin.haller@anu.edu.au<mailto:armin.haller@anu.edu.au>>; Krzysztof Janowicz <jano@geog.ucsb.edu<mailto:jano@geog.ucsb.edu>>; Joshua Lieberman <jlieberman@tumblingwalls.com<mailto:jlieberman@tumblingwalls.com>>; Cox, Simon (L&W, Clayton) <Simon.Cox@csiro.au><mailto:Simon.Cox@csiro.au>; Kerry Taylor <kerry.taylor@anu.edu.au<mailto:kerry.taylor@anu.edu.au>>
Cc: SDW WG Public List <public-sdw-wg@w3.org<mailto:public-sdw-wg@w3.org>>
Subject: Re: rdfs:class versus owl:class in SOSA-Core

On 09/11/2016 23:46, Krzysztof Janowicz wrote:


>> -rdfs:domain and rdfs:range
>> Please use these whenever relevant. [...]. You can't go wrong with them.
> This is the only part of your email where I would strongly disagree
> (well, depending on what you mean by 'whenever relevant' :-)) for the
> reasons we discussed here before and that others made when developing
> other ontologies. This is not to say that using them is 'wrong' in
> some sense but that they have to be handled with great care.

On this, I know what you mean. I've read the arguments that were written at some point in the SSN draft. I know that it is wise to avoid systematic domains and ranges that could be too restrictive.

But I think that this went too far with things like:

:hasMeasurementCapability  a  owl:ObjectProperty .
:MeasurementCapability  a  owl:Class .

and no range for :hasMeasurementCapability. If, whenever I find:

  ?x :hasMeasurementCapability  ?y .

  I cannot conclude that ?y is a :MeasurementCapability, then there is something really wrong in the name of the property!  Or, alternatively there ought to be a rdfs:range axiom.


> Best,
> Krzysztof
> On 11/09/2016 01:15 PM, Antoine Zimmermann wrote:
>> On 09/11/2016 00:20, Armin Haller wrote:
>>> Hi Krzysztof,
>>> Thanks for your detailed explanations below!
>>> Just to clarify, the intention in the meeting to go through a list
>>> of what constructs should be in SOSA (as thankfully proposed by
>>> Josh) was to be incremental. I was planning to incrementally go
>>> through the list of constructs that are either already in our
>>> current SOSA proposal or could be imagined to be in it and vote on
>>> them. Some, of course have implications, if we decide on
>>> owl:inverseOf in our next meeting, we will not be in RDFS entailment.
>> I was not at the meeting, so I may have missed something, but what is
>> the rationale for forbidding certain constructs?  If there is an
>> owl:inverseOf in the ontology, the RDFS reasoners won't care. Nobody
>> who's not using/interested in owl:inverseOf will care.
>> However, in the case of owl:inverseOf, I don't think it is a question
>> of whether we want to use the construct or not. It is a question of
>> deciding whether we allow ourselves to define both a property and its
>> inverse. If we have properties like ex:hasChild and ex:hasParent, it
>> would be silly not to make explicit the inverse relationship. If the
>> decision is about not having both a property and its inverse, then
>> the need to use or not owl:inverseOf is only an inevitable
>> consequence of the other decision.
>>> If we are already in OWL, then of course it would make sense to use
>>> owl:Class, although we do not have to. Therefore, again a vote on
>>> owl:Class thereafter.
>> I don't see a reason not to. Not having owl:Class declaration has
>> noticeable consequences. For instance, it always troubles me that
>> Protégé cannot display the Dublin Core properties and classes. It
>> also troubles me that the DC terms cannot be imported with standard
>> owl:imports, because of the lack of owl class, owl properties
>> declarations.
>>> I can think of the following list to vote on in our next meeting,
>>> incrementally. And we stopped at owl:inverseOf this meeting I just
>>> saw in the minutes.
>>> -rdfs:class
>> see above and related emails.
>>> -owl:inverseOf
>> see above.
>>> -owl:AnnotationProperty
>>> -           owl:ObjectProperty
>> owl:AnnotationProperty, owl:ObjectProperty and owl:DatatypeProperty,
>> like owl:Class, do not add expressiveness to the language. However,
>> they help OWL tools figuring out what the terms are.
>> FWIW, in OWL 2 RDF-based semantics, owl:ObjectProperty is equivalent
>> to rdf:Property. owl:AnnotationProperty and owl:DatatypeProperty are
>> both subClassOf rdf:Property.
>>> -owl:Class
>> see above.
>>> -rdfs:domain and rdfs:range
>> Please use these whenever relevant. When used with an atomic class,
>> they are fully supported by RDFS, OWL 2 EL, OWL 2 QL, OWL 2 RL, OWL 2
>> DL, OWL 2 Full, ter Horst semantics, RDFS++. You can't go wrong with
>> them.
>>> -rdfs:subClassOf
>> Come on! Do we need to vote on this one?
>>> -           owl:Restriction
>> There are many forms of restrictions, each one should be considered
>> individually. I don't see much reason to forbid ourselves to use any
>> of them because anyone can just ignore those they don't like.
>> However, they should be used sensibly because we should not prevent
>> use cases that we have not foreseen. I'd prefer an ontology that has
>> very little restrictions with precise documentation to prevent
>> misusing the terms, rather than an overly restrictive one. Look at
>> schema.org<http://schema.org>: painfully non-restrictive, delightfully useful.
>> --AZ
>>> Please do think about these and if you think they should or should
>>> not be in the core or if there is anything else we desperately would need.
>>> Kind regards,
>>> Armin
>>> *From: *Krzysztof Janowicz <jano@geog.ucsb.edu<mailto:jano@geog.ucsb.edu>>
>>> *Date: *Wednesday, 9 November 2016 at 10:01 am
>>> *To: *Joshua Lieberman <jlieberman@tumblingwalls.com<mailto:jlieberman@tumblingwalls.com>>, Simon Cox
>>> <Simon.Cox@csiro.au><mailto:Simon.Cox@csiro.au>, Kerry Taylor <kerry.taylor@anu.edu.au<mailto:kerry.taylor@anu.edu.au>>, Armin
>>> Haller <armin.haller@anu.edu.au<mailto:armin.haller@anu.edu.au>>
>>> *Cc: *SDW WG Public List <public-sdw-wg@w3.org<mailto:public-sdw-wg@w3.org>>
>>> *Subject: *rdfs:class versus owl:class in SOSA-Core
>>> Hi,
>>> Sorry for being so picky about this during our meeting but I do not
>>> want us to take decisions that have consequences that we can not yet foresee.
>>> To the best of my knowledge (and please correct me if I am wrong):
>>> Under the semantics of OWL1, rdfs:class and owl:class are only
>>> equivalent for OWL-Full. For OWL-DL (and OWL-Lite) owl:class is a
>>> subclass of rdfs:class.
>>> This means that every valid document in OWL will be a valid document
>>> in RDFS, however *not* every rdfs:class is an owl:class. I do not
>>> want us to end up in OWL-Full because of this.
>>> For OWL2, I found this: 'owl:Class rdfs:subClassOf rdfs:Class . "
>>> (https://www.w3.org/TR/owl2-rdf-based-semantics/). Things may be
>>> more complicated here due to OWL2 punning and they may well turn out
>>> to be equivalent, I will check this later.
>>> If we decide to restrict ourself to only using RDFS for SOSA-core,
>>> and I am not in favor of this, then we may have to go with rdfs:class.
>>> However, we have not yet taken this decision and have also not
>>> discussed which axioms and language to use for SSN. As Sosa-core and
>>> SSN will be aligned, this may have more consequences that we should
>>> consider. It also seems like many of us are in favor of using
>>> inverseOf, so we would be using OWL (and its formal semantics)
>>> anyway. Note that this does not do any harm to an RDFS-only
>>> tool/user as for those the inverseOf axiom will simply have no
>>> formal semantics. Still all other triples that use both relations will still be just fine.
>>> Given the subclasssing, I do not see any problems using owl:class,
>>> but we may accidentally end up in OWL-full or with being
>>> incompatible to the standards if we opt for rdfs:class. Again, I am happy to be corrected.
>>> At least, I do not see harm in simply using owl:class.
>>> Finally, and from very pragmatic point of view: ontologies that are
>>> under very heavy use such as the DBpedia ontology simply use
>>> owl:class and I have not yet seen any issues or complaints about that. See, for
>>> example, http://dbpedia.org/ontology/City "dbo:City    rdf:type
>>> owl:Class ." The same is true for the goodrelations ontology and so
>>> forth (but I admit that this is due to the more complex
>>> axiomatization they use).
>>> I hope this will start a productive discussion.
>>> Thanks for reading,
>>> Krzysztof


Krzysztof Janowicz

Geography Department, University of California, Santa Barbara

4830 Ellison Hall, Santa Barbara, CA 93106-4060

Email: jano@geog.ucsb.edu<mailto:jano@geog.ucsb.edu>

Webpage: http://geog.ucsb.edu/~jano/

Semantic Web Journal: http://www.semantic-web-journal.net

Received on Saturday, 12 November 2016 05:21:34 UTC

This archive was generated by hypermail 2.4.0 : Thursday, 24 March 2022 20:31:27 UTC