playing around with Renaud's term search stuff, I find some quite interesting stuff there...
Axel Polleres
e.g. looking for "location" and "foaf" I find several mentions of the non-existant property
foaf:location ... e.g. http://myopenlink.com:8890/DAV/home/sdmonroe/tiwan.rdfs
which is interesting... looks like that term-search tool, if we'd tidy it up a bit wouldn't only potentially serve for finding commonly used ontology terms, but also erroneous usage of terms...
Axel
Antoine Zimmermann
- From an RDF-semantic point of view, a property exists as soon as it
is used as a property.
- From a DL point of view, a property exists as soon as soon as it is
defined as iether an ObjectProperty, a DatatypeProperty or an
AnnotationProperty.
The property http://xmlns.com/foaf/0.1/location satisfies both.
From a linked data perspective, the only problem is that this URI does
not dereference to a document that describes it.
2010/2/18 Axel Polleres <axel.p...@deri.org>:
--
--AZ
Dan Brickley
<antoine.z...@gmail.com> wrote:
> Depends what you mean by "non-existant".
> - From an RDF-semantic point of view, a property exists as soon as it
> is used as a property.
> - From a DL point of view, a property exists as soon as soon as it is
> defined as iether an ObjectProperty, a DatatypeProperty or an
> AnnotationProperty.
Neither the RDF nor OWL (DL or otherwise) semantics specs really get
into the notion of authority that comes with our use of URIs to name
things. They don't model change either. That stuff is both important,
and really quite hard to characterise using the mathematical
formalisms that define the core of RDF/RDFS/OWL. Somehow we survive
this tension.
If I write http://purl.org/dc/terms/author as claim it is a property,
I am simply mistaken; it isn't (currently or formerly).
The way we grew the FOAF vocab in early days was grassroots and
deployment led, which is what led to a somewhat messy collection of
property names. These days we have metadata within the spec (in html
and rdf) to indicate how experimental a property is.
> The property http://xmlns.com/foaf/0.1/location satisfies both.
There is no such property, yet. I don't think it's particularly
helpful to propagate the view that things in other people's namespaces
can be 'wished into being'. However it is helpful to build
infrastructure that lets people throw around ideas informally before
they end up standardised; so it is quite possible 'location' will end
up being defined in the FOAF spec someday. And if there's a lot of
data out there already using it, we might even take that into account
(it would be foolish not to :).
> From a linked data perspective, the only problem is that this URI does
> not dereference to a document that describes it.
No, there are lots of information-linking problems when people make
things up without documenting them. The lack of a direct link to the
documentation is only one. Is foaf:location a functional property
(owl:FunctionalProperty), for example? What kinds of things does it
apply to, and what kinds of values does it take? Can a thing be in two
places at once? Can some things never be in a place? etc...
cheers,
Dan
Antoine Zimmermann
2010/2/18 Dan Brickley <dan...@danbri.org>:
> On Thu, Feb 18, 2010 at 4:41 PM, Antoine Zimmermann
> <antoine.z...@gmail.com> wrote:
>> Depends what you mean by "non-existant".
>> - From an RDF-semantic point of view, a property exists as soon as it
>> is used as a property.
>> - From a DL point of view, a property exists as soon as soon as it is
>> defined as iether an ObjectProperty, a DatatypeProperty or an
>> AnnotationProperty.
>
> Neither the RDF nor OWL (DL or otherwise) semantics specs really get
> into the notion of authority that comes with our use of URIs to name
> things. They don't model change either. That stuff is both important,
> and really quite hard to characterise using the mathematical
> formalisms that define the core of RDF/RDFS/OWL. Somehow we survive
> this tension.
I'm not sure what you are arguing for or against, but what I said has
nothing to do with authority. I just said that, from an RDF-(or
RDFS-)perspective, any URI that is used as a property (e.g., in the
predicate position of a triple; or as the subject of a triple ?p
rdfs:range ?C) is interpreted as a property, and therefore, to a
certain extent, exists. This is the case of foaf:location in document
[1].
From a OWL DL perspective, a property exists iff it is declared as one
of the three ObjectProperty, DatatypeProperty or AnnotationProperty
[2]. The URI foaf:location in [1] is declared as an ObjectProperty,
and therefore, in this sense, "exists".
From a formal semantics point of view, the fact that foaf:location
starts with the same sequence of characters as any other FOAF
properties is purely coincidental. This is a limitation of the
RDF/RDFS/OWL specs.
[1] http://myopenlink.com:8890/DAV/home/sdmonroe/tiwan.rdfs
[2] 5.8 Entity Declarations and Typing. In OWL 2 Web Ontology Language
- Structural Specification and Functional-Style Syntax.
http://www.w3.org/TR/owl2-syntax/#Entity_Declarations_and_Typing
> If I write http://purl.org/dc/terms/author as claim it is a property,
> I am simply mistaken; it isn't (currently or formerly).
Well, I don't see why not. If it pleases you that this is a property
and you do interesting things with it, then let it be a property.
Nonetheless, not everyone would agree and it may be detrimental in
other contexts. That's wy we have to build systems that are robust to
the heterogeneity of descriptions. You can't avoid people saying
mutually inconsistent things but you can avoid your own system
crashing each times it happens.
> The way we grew the FOAF vocab in early days was grassroots and
> deployment led, which is what led to a somewhat messy collection of
> property names. These days we have metadata within the spec (in html
> and rdf) to indicate how experimental a property is.
This is fine and good (the metadata). Actually, it would be good to
have a standard (or "de facto" standard) vocabulary to annotate
ontological terms. That would in particular help deciding whether a
description is trustable based on its use of deprecated or unstable
terms.
>> The property http://xmlns.com/foaf/0.1/location satisfies both.
>
> There is no such property, yet. I don't think it's particularly
> helpful to propagate the view that things in other people's namespaces
> can be 'wished into being'. However it is helpful to build
> infrastructure that lets people throw around ideas informally before
> they end up standardised; so it is quite possible 'location' will end
> up being defined in the FOAF spec someday. And if there's a lot of
> data out there already using it, we might even take that into account
> (it would be foolish not to :).
Well, there *is* such property in document [1]. It is simply not in
the FOAF spec. If the authors of [1] are doing useful things with
foaf:location, then there is a property that should not be ignored.
However, we, the Pedantic Web Group, should inform them that their use
of the foaf: prefix is going against good practices, and they should
change it. But still, if they don;t want to change it and do useful
things with it, it should not be ignored, in spite of it being a kind
of "prefix-hijacking".
>> From a linked data perspective, the only problem is that this URI does
>> not dereference to a document that describes it.
>
> No, there are lots of information-linking problems when people make
> things up without documenting them. The lack of a direct link to the
> documentation is only one. Is foaf:location a functional property
> (owl:FunctionalProperty), for example? What kinds of things does it
> apply to, and what kinds of values does it take? Can a thing be in two
> places at once? Can some things never be in a place? etc...
What I wanted to say here is that the problem with foaf:location is
not specific to properties or ontology terms. It does not respect a
basic Linked Data principle that we already take into account in our
recommended practices. So it is not a new problem, so to speak.
> cheers,
I hope I made myself clear enough.
Best,
AZ.
>
> Dan
>
>> 2010/2/18 Axel Polleres <axel.p...@deri.org>:
>>> ...
>>>
>>> e.g. looking for "location" and "foaf" I find several mentions of the non-existant property
>>> foaf:location ... e.g. http://myopenlink.com:8890/DAV/home/sdmonroe/tiwan.rdfs
>>> which is interesting... looks like that term-search tool, if we'd tidy it up a bit wouldn't only potentially serve for finding commonly used ontology terms, but also erroneous usage of terms...
>>>
>>> Axel
>>
>> --
>> --AZ
>>
>
--
--AZ
Axel Polleres
On 18 Feb 2010, at 15:41, Antoine Zimmermann wrote:
> Depends what you mean by "non-existant".
> - From an RDF-semantic point of view, a property exists as soon as it
> is used as a property.
> - From a DL point of view, a property exists as soon as soon as it is
> defined as iether an ObjectProperty, a DatatypeProperty or an
> AnnotationProperty.
>
- From the best practices POV, it should be dereferenceable, and some information about it should be found there...
i call it non-existant, if this is not followed (particularly, this doesn't work, if someone "defines" a URI outside
her/his control... (which is basically what we call "speaking non-authoritatively about that URI" in SAOR, BTW)
> The property http://xmlns.com/foaf/0.1/location satisfies both.
>
> From a linked data perspective, the only problem is that this URI does
> not dereference to a document that describes it.
That IS the problem, yes, doesn't follow best practices.