Can we use full URLs for CRS or abbreviations?

[joanma747]

Features CRS extension uses Full CRS but we are using abbreviations. It seems that full URLS are considered ids. It might be better to use full URLs instead

[jerstlouis]

The OGC NA policy on short-hand CURIEs forms of CRS URIs was applied this week ( opengeospatial/NamingAuthority#92 and opengeospatial/NamingAuthority#93 ).

See https://docs.opengeospatial.org/pol/09-048r6.html#toc23 .

We should include text explicitly allowing CURIEs as short-hand for full URIs in OGC API - Maps.

[jerstlouis]

The policy describing the use of CURIEs is at https://docs.opengeospatial.org/pol/09-048r6.html#toc14 .
Currently it mentions:

OGC Standards should only allow safe CURIEs, unless support for other forms of CURIEs is inherited from another standard

Not sure what the value of safe CURIE is if there is no way that a non-safe CURIE could be misinterpreted?

Safe CURIE would force us to use:

https://example.org/foo/collections/bar/map?crs=[EPSG:3395]

instead of

https://example.org/foo/collections/bar/map?crs=EPSG:3395

[jerstlouis]

It seems that the requirements in the specification should be that the server shall support safe CURIEs (e.g., crs=[EPSG:4326]), but we should include a permission that the server may also be tolerant and interpret unsafe CURIE, which would be compatible with the WMS crs=EPSG:4326 syntax.

[joanma747]

I'm not sure that adding a "[" "]" makes a string more safe :-). In any case I'm starting to see this very confusing. Too many notations for the same thing only because we are trying to please too many people: the old WMS users, the NA people and the RDF people.

[jerstlouis]

@joanma747

The square brackets may be used to prevent ambiguities between CURIEs and regular URIs, yielding so-called safe CURIEs.

from https://en.wikipedia.org/wiki/CURIE

See also https://www.w3.org/TR/curie/#sec_2.2 2.2. Ambiguities Between CURIEs and URIs

Too many notations for the same thing only because we are trying to please too many people:

Yes, the requirements, based on the policy, would express that servers shall support the two following options:

• http://www.opengis.net/def/crs/EPSG/0/4326
• [EPSG:4326]

and a permission would allow the server implementations to also recognize simply EPSG:4326, making the crs query parameter compatible with WMS.

I feel it is good if we can make everybody happy ;)

[chris-little]

Can EPSG:4326 cause wrong HTTP and namespace behaviour, outside of a geospatial context?

[jerstlouis]

@chris-little I don't believe so currently, but as @cportele points out in opengeospatial/ogcapi-features#632 (comment) new URI schemes are added all the time.

[chris-little]

@jerstlouis My preference is for the Naming Authority policy (safe CURIEs) to remain. I think putting square brackets is not burdensome, is simple, and clearly avoids ambiguity. I see many examples of geospatial software using them, even doubled, to avoid ambiguity in a WKT context, such as optionally supporting arrays rather than single values.
I am sympathetic to syntaxes and notation being global, generic and non-domain specific.
Safe CURIEs like [EPSG:nnnn]` is apolicy, so I suppose implementations are free to choose to be more generous, but I think Postel's Rule applies: accept generously, generate strictly.

[joanma747]

Applied and closed in regular telco