Handling of sioc:User deprecation in vocabulary

[Hogan, Aidan]

Hi,

Just some feedback with respect to reflecting the recent deprecation of
sioc:User in the vocabulary [1]. It seems that you define an
equivalentClass relation between the deprecated term sioc:User and the
new term sioc:UserAccount. This means that every member of the new
sioc:UserAccount class will infer a member of the deprecated sioc:User
class. In my opinion, sioc:User should be removed from all "inference
paths" involving non-deprecated terms. I would rather favour a subclass
relation from sioc:User to sioc:UserAccount (i.e., infer recommended
terms from deprecated terms, not the other way around).

More arguably, I would consider changing the domain of deprecated
properties such as sioc:last_name from sioc:User to sioc:UserAccount.
This is more debatable however.

Cheers,
Aidan
- http://pedantic-web.org/

[1] http://rdfs.org/sioc/ns#

--
You received this message because you are subscribed to the Google Groups "SIOC-Dev" group.
To post to this group, send email to sioc...@googlegroups.com.
To unsubscribe from this group, send email to sioc-dev+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/sioc-dev?hl=en.

[Richard Cyganiak]

On 19 May 2010, at 00:02, Hogan, Aidan wrote:
> In my opinion, [deprecated class] sioc:User should be removed from

> all "inference
> paths" involving non-deprecated terms. I would rather favour a
> subclass
> relation from sioc:User to sioc:UserAccount (i.e., infer recommended
> terms from deprecated terms, not the other way around).

I'd say that a smart reasoner should have an optional setting that
stops it from inferring assertions involving deprecated classes and
properties.

After all, sioc:User and sioc:UserAccount *are* equivalent in terms of
OWL semantics. It's just that the use of the one class is discouraged.
The current state of the SIOC ontology reflects this intention, IMO.

> More arguably, I would consider changing the domain of deprecated
> properties such as sioc:last_name from sioc:User to sioc:UserAccount.

+1 on this one. The properties are deprecated; that doesn't mean they
should be excluded from maintenance.

Best,
Richard

[Alexandre Passant]

On 19 May 2010, at 15:36, Richard Cyganiak wrote:

>
> On 19 May 2010, at 00:02, Hogan, Aidan wrote:
>> In my opinion, [deprecated class] sioc:User should be removed from all "inference
>> paths" involving non-deprecated terms. I would rather favour a subclass
>> relation from sioc:User to sioc:UserAccount (i.e., infer recommended
>> terms from deprecated terms, not the other way around).
>
> I'd say that a smart reasoner should have an optional setting that stops it from inferring assertions involving deprecated classes and properties.
>
> After all, sioc:User and sioc:UserAccount *are* equivalent in terms of OWL semantics. It's just that the use of the one class is discouraged. The current state of the SIOC ontology reflects this intention, IMO.

I agree with Richard - unless I missed something with your comment, Aidan.
And I think that adding it as a subclass will change its semantics (they are the same thing, User is not more specific than UserAccount)

>
>> More arguably, I would consider changing the domain of deprecated
>> properties such as sioc:last_name from sioc:User to sioc:UserAccount.
>
> +1 on this one. The properties are deprecated; that doesn't mean they should be excluded from maintenance.

I've edited the spec. file in SVN

Alex.

>
> Best,
> Richard
>
>
>
>> This is more debatable however.
>>
>> Cheers,
>> Aidan
>> - http://pedantic-web.org/
>>
>> [1] http://rdfs.org/sioc/ns#
>>
>
> --
> You received this message because you are subscribed to the Google Groups "SIOC-Dev" group.
> To post to this group, send email to sioc...@googlegroups.com.
> To unsubscribe from this group, send email to sioc-dev+u...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/sioc-dev?hl=en.
>

--
Dr. Alexandre Passant
Digital Enterprise Research Institute
National University of Ireland, Galway
:me owl:sameAs <http://apassant.net/alex> .

[Hogan, Aidan]

> >> In my opinion, [deprecated class] sioc:User should be removed from
all
> "inference
> >> paths" involving non-deprecated terms. I would rather favour a
subclass
> >> relation from sioc:User to sioc:UserAccount (i.e., infer
recommended
> >> terms from deprecated terms, not the other way around).
> >
> > I'd say that a smart reasoner should have an optional setting that
stops
> it from inferring assertions involving deprecated classes and
properties.
> >
> > After all, sioc:User and sioc:UserAccount *are* equivalent in terms
of
> OWL semantics. It's just that the use of the one class is discouraged.
The
> current state of the SIOC ontology reflects this intention, IMO.
>
> I agree with Richard - unless I missed something with your comment,
Aidan.
> And I think that adding it as a subclass will change its semantics
(they
> are the same thing, User is not more specific than UserAccount)

Yes, [1] sets a precedent backing up Richard and Alexandre's position.

PROS for using equivalentClass (not using subClassOf):
* sioc:UserAccount will not be considered deprecated;
* maintains better backwards-compatibility with systems expecting
sioc:User memberships to be inferred as before;
* the claim is perfectly valid;
* precedent in a standard.

CONS for using equivalentClass (not using subClassOf):
* redundancy in terms, where you now effectively have two terms
identifying the same extension (members);
* all members of sioc:UserAccount will be inferred to be a member of a
deprecated class in all standard/legacy reasoning tools;
* I have to fix the RDF:Alerts system to account for such modelling.

Essentially, the former approach leads to the introduction of a
synonymous preferred term for sioc:User with strong support for the
deprecated term, and the latter approach leads towards the exclusion of
sioc:User from reasoning and shows more intent to completely phase out
the term.

<unnecessary-segue skip="optional">
What is also becoming more clear to me is two different philosophies for
RDFS/OWL:

* a 'claim-based' philosophy for describing terms in RDFS/OWL, in which
you state what you hold to be true for that term, modelling the
important aspects of your view of usage of the term;
* an 'inferencing-based' philosophy for describing terms in RDFS/OWL, in
which you create descriptions with a specific eye to creating useful
inferences through those terms.

...and, of course, various combinations of the two. I tend heavily
towards the latter (possibly in minority); for example, I would probably
never bother defining the domain of a property to be rdfs:Resource -- no
useful inference will come out of it -- whereas some people would see
that as making the claim *you can use the property on any kind of
subject*.

I think using 'equivalentClass' here tends towards a 'claim-based' view
of OWL/RDFS, whereas using 'subClassOf' would probably be more practical
from an inferencing perspective.
</unnecessary-segue>

Cheers,
Aidan

[1] http://www.w3.org/TR/owl-ref/#Deprecation