W3C home > Mailing lists > Public > public-sdw-wg@w3.org > May 2017

RE: SpatialThing and feature (again)

From: <Simon.Cox@csiro.au>
Date: Wed, 3 May 2017 01:31:56 +0000
To: <jeremy.tandy@gmail.com>, <portele@interactive-instruments.de>
CC: <matthew.perry@oracle.com>, <public-sdw-wg@w3.org>
Message-ID: <9fac75809059428c84305358fc15354d@exch1-mel.nexus.csiro.au>
And I’ve provided some figures to help.

From: Jeremy Tandy [mailto:jeremy.tandy@gmail.com]
Sent: Wednesday, 3 May, 2017 07:01
To: Clemens Portele <portele@interactive-instruments.de>
Cc: Matthew Perry <matthew.perry@oracle.com>; public-sdw-wg@w3.org
Subject: Re: SpatialThing and feature (again)

Hi - I've added a green note the BP doc (at end of §5) via PR 781 [1]. See the commit [2] if you want to know the details :-)


[1]: https://github.com/w3c/sdw/pull/781

[2]: https://github.com/w3c/sdw/commit/89241e63b7eab37bce1f62e608dc78b2d63b3961

On Tue, 2 May 2017 at 17:20 Jeremy Tandy <jeremy.tandy@gmail.com<mailto:jeremy.tandy@gmail.com>> wrote:
Thanks Clemens.
On Tue, 2 May 2017 at 17:18, Clemens Portele <portele@interactive-instruments.de<mailto:portele@interactive-instruments.de>> wrote:
Hi Jeremy,

I see your point and I guess you are right. I still think that the description is quite fuzzy and it is not obvious that "anything with spatial extent" is not meant to be narrower than the ISO 19109 concept of a feature *), which is partly why I started the thread. But the text is quite challenging for new readers already due to all the legacy of terms and definitions that are not well aligned and adding more nuanced discussion is probably not helpful. So I am ok with leaving it as it is. We always reference back to the email thread ;)


*) I am ignoring here for simplicity that spatial thing includes not only the abstraction, but also the real-world entity.

On 2. May 2017, at 17:52, Jeremy Tandy <jeremy.tandy@gmail.com<mailto:jeremy.tandy@gmail.com>> wrote:

Hi - I'm just in the process of updating the BP document to reflect our discussion.

Clemens suggested that we explicitly call out things like "home loan" as an example of a spatial thing. Having read through the text, it feels like this is a fairly nuanced statement that A/ may lead to more confusion, unless B/ we take time to describe why something that appears to not have spatial extent really does.

Personally, I'd rather leave this complexity out.

What do you think? (especially Clemens)


On Tue, 25 Apr 2017 at 12:58 Jeremy Tandy <jeremy.tandy@gmail.com<mailto:jeremy.tandy@gmail.com>> wrote:
Thanks all. I can amend the BP doc to clarify as per Simon's proposal. Jeremy
On Tue, 25 Apr 2017 at 12:54, Matthew Perry <matthew.perry@oracle.com<mailto:matthew.perry@oracle.com>> wrote:

That looks correct to me as well.


On 4/25/2017 12:29 AM, Joshua Lieberman wrote:
Yes, I think.

On Apr 25, 2017, at 12:19 AM, <Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au>> <Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au>> wrote:

>  ... and reaffirm that _we_ see Feature (SpatialThing) as disjoint from geometry, but that this might be at odds with some people's interpretations. As Josh says: "we can’t really say there is a mapping from W3C Basic Geo to/from anything based on 19109."

We need to be very clear here:

geosparql:SpatialObject         includes both features and geometries – they are disjoint subclasses
w3cgeo:SpatialThing               is superclass of w3cgeo:Point, but (OWA) potentially also has a class of features as another subclass (disjoint from Point) – so this could all be OK and consistent (but we mustn’t credit w3cgeo as having been the result of much deep thought).

So where does bp:SpatialThing fit in? Looks to me like the key thing is to point out that it is *not* the same as w3cgeo:SpatialThing, because the latter includes geometries. But it *is* the same as geosparql:Feature, which is disjoint from Geometry.


From: Clemens Portele [mailto:portele@interactive-instruments.de]
Sent: Tuesday, 25 April, 2017 01:27
To: Jeremy Tandy <jeremy.tandy@gmail.com<mailto:jeremy.tandy@gmail.com>>
Cc: Josh Lieberman <jlieberman@tumblingwalls.com<mailto:jlieberman@tumblingwalls.com>>; Cox, Simon (L&W, Clayton) <Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au>>; Rob Atkinson <rob@metalinkage.com.au<mailto:rob@metalinkage.com.au>>; Andreas Harth <harth@kit.edu<mailto:harth@kit.edu>>; public-sdw-wg@w3.org<mailto:public-sdw-wg@w3.org>
Subject: Re: SpatialThing and feature (again)

Hi Jeremy,

I think we should add a green note in chapter 5 to explain how the "anything with spatial extent" definition is consistent with features like a "home loan" in a spatial dataset as it is not obvious.


On 21. Apr 2017, at 17:33, Jeremy Tandy <jeremy.tandy@gmail.com<mailto:jeremy.tandy@gmail.com>> wrote:

Hi all-

I've spent more than a few minutes parsing through the email chain.

1/ Clemens' summary (from mid way though) suggests that (a) ISO 19109 Feature is [also] a geosparql:Feature, (b) these may or may not have attached geometry properties
2/ Andrea suggests that "only [those] ISO 19109 Features [with spatial extent] are Spatial Things according to the BP definition" - but Josh suggests we're using "spatial extent" as a shorthand for "real-world phenomena", and that "making the connection [between abstraction and real-world thing] formal and explicit is not necessary for Web purposes"

So I'm seeing that there's no inconsistency to explain away.

Please confirm that I've read this OK. Apologies if I've missed the point!

And, talking of Points ... I see that there is potential for confusion regarding the "Feature/Geometry amalgam".

We could insert a "green note" into the BP document identifying the potential for inconsistency - as defined in Andreas' example:

> Because a w3cgeo:SpatialThing has lat/lon, some people might equate a w3cgeo:SpatialThing with a geosparql:Geometry.
> Because the w3cgeo:SpatialThing is an instance of foaf:Person, some other people find it natural to equate the w3cgeo:SpatialThing with a geosparql:Feature.
> Based on data from different source we now have an inconsistency, because the w3cgeo:SpatialThing is an instance of both geosparql:Feature and geosparql:Geometry, which are defined as disjoint.

... and reaffirm that _we_ see Feature (SpatialThing) as disjoint from geometry, but that this might be at odds with some people's interpretations. As Josh says: "we can’t really say there is a mapping from W3C Basic Geo to/from anything based on 19109."

Am I summarising correctly?

Thanks, Jeremy

On Fri, 21 Apr 2017 at 15:33 Joshua Lieberman <jlieberman@tumblingwalls.com<mailto:jlieberman@tumblingwalls.com>> wrote:
Ah, I had thought that the domains of geo:lat and geo:lon were geo:Point, since that is what is generally referred to in narrative. If a resource carrying the lat/lon properties implies that it is a SpatialThing, not only the Point subclass, adding the properties doesn’t resolve any feature / geometry ambiguity. Your equivalences are certainly possible, but geosparql doesn’t / shouldn’t support adding direct positions to features, so entailing something with geo:lat and geo:lon as geosparql:SpatialObject rather than geosparql:Geometry doesn’t really work. And if we can’t derive that use of geo:lat and geo:lon imply both a feature and a geometry, than Andrea is correct that we can’t really say there is a mapping from W3C Basic Geo to/from anything based on 19109. That may be unfortunate.


On Apr 20, 2017, at 8:38 PM, simon.cox@csiro.au<mailto:simon.cox@csiro.au> wrote:

Hold on a moment folk – does this problem really exist?

I’m looking at http://www.w3.org/2003/01/geo/wgs84_pos#<http://www.w3.org/2003/01/geo/wgs84_pos> which is the RDF/XML serialization of W3C Basic Geo.
Here’s the key axioms.

geo:lat   rdfs:domain geo:SpatialThing .
geo:long rdfs:domain geo:SpatialThing .
geo:Point rdfs:subClassOf geo:SpatialThing .

And from http://schemas.opengis.net/geosparql/1.0/geosparql_vocab_all.rdf


geosparql:Geometry rdfs:subClassOf geosparql:SpatialObject .

then it looks to me like

geo:SpatialThing owl:equivalentClass geosparql:SpatialObject .
geo:Point rdfs:subClassOf geosparql:Geometry .

and there is no inconsistency. Appearance of geo:lat and geo:long properties only entails that it is a geosparql:SpatialObject, so can be either a Feature or a Geometry.

Am I missing something?


From: Rob Atkinson [mailto:rob@metalinkage.com.au]
Sent: Thursday, 20 April, 2017 06:24
To: Joshua Lieberman <jlieberman@tumblingwalls.com<mailto:jlieberman@tumblingwalls.com>>; Andreas Harth <harth@kit.edu<mailto:harth@kit.edu>>
Cc: public-sdw-wg@w3.org<mailto:public-sdw-wg@w3.org>
Subject: Re: SpatialThing and feature (again)

This could also be resolved by thinking of geo:long as a property that can entail a geometry property of the feature - maybe its even a geometry property in the same way that a 2D point is a partial representation of a 3D location?


On Thu, 20 Apr 2017 at 02:38 Joshua Lieberman <jlieberman@tumblingwalls.com<mailto:jlieberman@tumblingwalls.com>> wrote:

It may not be worth delving too deeply into this...

W3C Basic Geo defines SpatialThing and then subclasses it to Point carrying the lat and long properties. No one defines their own SpatialThings, they simply add geo:lat and geo:long properties to some resource X to turn it into “also a Point”, in other words “also a geometry”. This implies for most users but does not actually assert that resource X is both a feature and a geometry. One could form a subclass of geo:SpatialThing that was actually disjoint with geo:Point or other geometry,  which would then align more-or-less with iso geosparql:Feature, hence the assertion that some geo:SpatialThings are geosparql:Features. This is largely hypothetical.

There is a similar property in GeoRSS, the point(pos) property, but this doesn’t try to create one feature-geometry amalgam. It’s simply a shortcut for a longer expression that identifies some resource as a _Feature with a “where" object property connecting to a Point geometry resource.

It might be most accurate to say that your example of using W3C Basic Geo to represent feature and geometry in the “style” of geosparql is actually the longhand of what people are trying to do when they do use geo:lat and geo:long, identifying a resource as a real world feature and giving it a closely allied point geometry.


> On Apr 19, 2017, at 11:54 AM, Andreas Harth <harth@kit.edu<mailto:harth@kit.edu>> wrote:
> Hi,
> On 04/19/17 13:29, Joshua Lieberman wrote:
>> My understanding based on the limited documentation is that w3cgeo:SpatialThing covers both features and models such as geometries, so
> that's my understanding too.  With the W3C WGS84 vocabulary you can write:
> @prefix geo: <http://www.w3.org/2003/01/geo/wgs84_pos#<http://www.w3.org/2003/01/geo/wgs84_pos>> .
> @prefix : <#> .
> :bob a geo:SpatialThing ; geo:lat "52.5196143" ; geo:long "13.4065603" .
> So the resource with the URI :bob is both the "feature" and the "geometry".
> In other representations (NeoGeo, GeoSPARQL), you would identify two separate
> resources:
> @prefix geo: <http://www.w3.org/2003/01/geo/wgs84_pos#<http://www.w3.org/2003/01/geo/wgs84_pos>> .
> @prefix : <#> .
> :bob a :Feature ; :geometry _:bnode .
> _:bnode a :Geometry , geo:Point ; geo:lat "52.5196143" ; geo:long "13.4065603" .
> The URI :bob now represents the "feature" resource, and the blank node _:bnode
> represents the "geometry" resource.
> I wouldn't know how to write OWL axioms to map the two modeling choices though.
> Best regards,
> Andreas.

Received on Wednesday, 3 May 2017 01:32:49 UTC

This archive was generated by hypermail 2.4.0 : Thursday, 24 March 2022 20:31:32 UTC