[Geojson] PR #199: CRS

[Martin Thomson <[e-mail]>]

Surprising no one, this issue turns out to be tricky.

If I understand the discussion this is the intended statements we want to make:

1. Using a different CRS is not advised.
2. Using a different CRS with lat, long ordering of coordinates is prohibited.
3. Anyone using a different CRS is responsible for ensuring that their
GeoJSON-like data is properly understood by all potential recipients.
They can do that by limiting the set of recipients.

The second point is one I'd like to get some clarity on.  It seems
that there is a fuzzy line here and I think we could be crisper about
where that line is.

I get that this particular error has caused interoperability problems,
but what about other sorts of confusion.  Is the problem specific to
an inversion of lat/lng?

Specifically, would the same sort of problem occur for a ECEF
Cartesian CRS?  Would the same problem occur for a CRS with a
dramatically different prime meridian (such as one that passes through
Mundubbera, or Vienna)?

If the former is a problem, but not the latter, then point 2 might
become "Using a CRS that uses a form other than long, lat or long,
lat, alt is prohibited."

Whereas if both issues are a concern, then it looks like we're
creating a rule that is based on some sort of subjective similarity to
WGS84.  Isn't it better to prohibit the use of alternative coordinate
reference systems entirely?  If we accept that, then people doing
their own thing, like Sean's UTM-based pipeline, are simply using
GeoJSON-like structures.