⇐ ⇒

[CF-metadata] Question on WKT representation of CRS

From: Bentley, Philip <philip.bentley>
Date: Mon, 3 Oct 2011 12:49:42 +0100

Hi Ethan,

All good questions! I've provided some brief - and probably inadequate -
responses inline.

I've been working up an initial (reasonably concise) proposal covering
this topic, which potentially I could upload to the CF trac site later
this week. I guess it depends on whether folks would prefer discussions
to happen against a dedicated trac ticket, or else here on the mailing
list?

Obviously if the preliminary sentiment is that use of WKT is a
non-starter, then that'll save me the trouble of submitting the proposal
:-)

Cheers,
Phil

> -----Original Message-----
> From: cf-metadata-bounces at cgd.ucar.edu
> [mailto:cf-metadata-bounces at cgd.ucar.edu] On Behalf Of Ethan Davis
> Sent: 01 October 2011 07:49
> To: CF metadata
> Subject: [CF-metadata] Question on WKT representation of CRS
>
> Hi Phil, all,
>
> This latest round of discussion around using CRS WKT reminded
> me of a few questions:
>
> 1) How easy is it to form a WKT that has good
> interoperability between different client libraries?
 
Not being familiar with the internals of the various libraries I'm not
well placed to judge. The authors of those libraries may wish to chime
in with their 2c worth.

> 2) Are there good specification documents for WKT?
>
> I believe the relevant OGC document is OGC 06-103r4 which can be found
> here: http://portal.opengeospatial.org/files/?artifact_id=25355

There's that definition. And also a more recent one in OGC document
01-009 -- Coordinate Transformation Services (at which point it became
known as CS WKT). As yet I've haven't really been able to establish
exactly who - if anyone - owns/maintains the WKT specification. From my
former days in the oil and gas exploration industry I recall it first
appearing under the (proprietary) auspices of POSC, long before open
source was de rigeur! But that consortium no longer exists AFAICT. Hence
it [the WKT syntax] seems to have just drifted into the public domain.
 
> But in the wild, there seem to be lots of variations in
> projection names and parameter names. There's even variation
> between them in the OGC spec mentioned above and the EPSG data base.

IMHO, some of the WKT examples in the OGC specs (and elsewhere) contain
errors, or at least deviations from the spec. Personally I consider the
EPSG (now OGP) registry to be the source of truth for CS/CRS entity
names and, where relevant, values. There is no other viable alternative
that I am aware of.
 
> 3) Given that, is CRS as WKT well enough defined so that
> providing a place for it in CF get us the interoperability we want?

I don't think anyone would claim that WKT, on it's own, represents an
interoperability silver bullet. But the received wisdom (as disseminated
on the CF mailing list!) seems to be that it would offer a very handy
adjunct to existing metadata descriptors.

> Or, Phil, would the proposal you have in mind go into more
> detail (or add to) the OGC spec?

That wasn't my intent.

> 4) One last question. The OGC CRS WKT specification is less
> than two pages of EBNF (without a lot of explanation). How
> far is the current CF grid mapping spec from complete
> coverage of the CRS WKT? Prime meridian and some units?

That I can't answer I'm afraid. As you know, there have been a number of
posts to the CF mailing list over recent months where people have stated
that having a WKT attribute would allow them to provide a more
complete/accurate definition of the coordinate system(s) used in their
netCDF files. I would hope that those same people would be able to
provide those examples as support for a putative CF proposal regarding
use of a WKT attribute.

Regards,
Phil
Received on Mon Oct 03 2011 - 05:49:42 BST

This archive was generated by hypermail 2.3.0 : Tue Sep 13 2022 - 23:02:41 BST

⇐ ⇒