[Geojson] GeoJSON 2.0

[Sean Gillies <[e-mail]>]

Hi all,

Getting to the finish requires us to reconcile two concerns:

1. We have consensus that WGS84 longitude and latitude only will maximize
interoperability.
2. The above is a breaking change (whether you agree the contract it breaks
was even working in the first place or not) and has to be done in a way
that does (ideally) no harm to implementers of geojson.org GeoJSON.

As Martin Thomson has suggested, and Dirk-Willem hinted at previously, I
propose we consider creating a GeoJSON version 2.0 to make the break while
giving implementers of the historical spec a unique name for what they're
implementing: "GeoJSON 1.0".

GeoJSON 1.0 would continue to be defined at
http://geojson.org/geojson-spec.html. It has no standard media type and no
standard file extension.

GeoJSON 2.0 would be our current draft with the change to "MUST NOT use CRS
other than the default". Its media type is "application/geo+json" and its
file extensions are ".geojson" and ".json". The media type
"application/geo+json" would always mean "GeoJSON 2.0".

I think it makes the most sense to, as previously suggested, add a
"version" member to the top-level object and give it a value of "2.0":

  {"type": "FeatureCollection", "version": "2.0", ... }

On receiving a GeoJSON document with no version member, a parser could try
to guess the version by looking for a "crs" member. If absent, the document
is practically equivalent to GeoJSON 2.0.

My apologies for the circular route we've taken to get here, but we've
certainly considered the alternatives to not versioning GeoJSON thoroughly
and have found them unacceptable.

-- 
Sean Gillies