R: Quick question after the Spatial Data on the Web Conference in Amsterdam

Da: Ed Parsons [mailto:eparsons@google.com]
Inviato: mercoledì 17 febbraio 2016 10:22
A: Erik Wilde <erik.wilde@dret.net>; Di Donato Pasquale swisstopo <Pasquale.DiDonato@swisstopo.ch>; public-sdw-comments@w3.org
Cc: Sean Gillies <sean.gillies@gmail.com>
Oggetto: Re: Quick question after the Spatial Data on the Web Conference in Amsterdam

>An interesting point, that was I think part of an ongoing more philosophical point we have discussed for the last few months.  The City Pasquale mentions as a thing of interest we would assign an identifier to which could be used to link to other objects containing information about >it, a record containing population, the name of the mayor, temperature forecast for today etc.

>Should we not think of the geometry representing the shape of the city in the same way, one record contains it geometry as modelling for small scale map production (a 1D point), another record the geometry associated with a midscale web mapping product (2d Polygon), we can >cope with a city having multiple football teams linking to the different information records for each team, why not different geometries ?

            Right, this is exactly my point. Within the GI community we see geometry just as a property of a feature (object), and a feature may have 0..* geometry-properties.
            In GeoJSON, as far as I understand it, geometry is a first class object and one cannot use more than one geometry-object, or am I wrong?

            Btw, many thanks for the nice discussion.

Pasquale


On Tue, 16 Feb 2016 at 09:45 Erik Wilde <erik.wilde@dret.net<mailto:erik.wilde@dret.net>> wrote:
hello pasquale.

On 2016-02-15 16:13 , Di Donato Pasquale swisstopo wrote:
> Here at swisstopo we are interested in GeoJSON-LD, since our SDI is
> entirely based on a REST API. Then the question about the “geometry”
> object in GeoJSON.

please keep in mind that there is no such thing as *the* "GeoJSON-LD".
the name has been used by various communities and generically refers to
some way how to use JSON-LD to map GeoJSON to RDF. how to best do this
is not trivial, in particular because GeoJSON relies so heavily on
sequences in its data model, which do not work well in RDF.

if and how there will be *a* GeoJSON-LD remains to be seen. in the IETF
GeoJSON WG (which i am co-chairing) we have had some discussions around
the idea, but for the upcoming RFC work we have decided that we are
focusing on GeoJSON only. GeoJSON-LD is out of scope.

> As far as I can understand the specifications, when I model a feature
> object I have to use the geometry object, with the restriction to only
> one geometry object.

https://tools.ietf.org/html/draft-ietf-geojson-00#section-2 says:

"GeoJSON always consists of a single object. This object (referred to as
the GeoJSON object below) represents a geometry, feature, or collection
of features."

so yes, GeoJSON by definition is a single JSON object. but you can have
a "Feature Collection" which is an array of Feature objects:

https://tools.ietf.org/html/draft-ietf-geojson-00#section-2.3


there also is a "Geometry Collection" which allows you to represent an
array of geometry objects:

https://tools.ietf.org/html/draft-ietf-geojson-00#section-2.1.8


> Is that right? Have you considered that sometimes geospatial features
> may have more than one geometry property? For example a city can be
> represented as a point or as a polygon depending on the scale of
> visualization.

the point of GeoJSON is not to represent the full complexity of spatial
models, for example different spatial representations on different zoom
levels. overloading GeoJSON this much very quickly will lead you to
GeoJSON that only you can meaningfully use, and which others will
interpret in different and unpredictable ways.

it makes more sense to think of GeoJSON as something that allows you to
represent geo features in a given context, such as a city on a certain
zoom level.

i am cc'ing sean gillies who is one of the original authors of GeoJSON
to make sure that i am not saying something that i shouldn't say. but
afaict, GeoJSON is intentionally modest and tries hard not to be the
spatial übermodel that can do it all.

cheers,

dret.

--
erik wilde | mailto:erik.wilde@dret.net<mailto:erik.wilde@dret.net> |
            | http://dret.net/netdret    |
            | http://twitter.com/dret    |
--

Ed Parsons
Geospatial Technologist, Google

Google Voice +44 (0)20 7881 4501
www.edparsons.com<http://www.edparsons.com> @edparsons