[gdal-dev] Re: [Proj] EPSG 7.9 Upgrade in Trunk

[Hamish]

Frank wrote:
> I have upgraded libgeotiff, PROJ.4 and GDAL with EPSG
> definitions regenerated from the EPSG 7.9 database.  New
> "preferred datum shift selection" rules are also in place which
> will give preference to 7 parameter datum shifts over
> 3 parameter datum shifts are also in place causing a fair
> amount of churn in the definitions.
> 
> The updates are in "trunk" for each of the projects. 
> I plan to produce libgeotiff and PROJ.4 release candidates
> soon with the new definitions so if you think you might have
> objections please investigate and speak up promptly.

Hi,

as per
  https://trac.osgeo.org/proj/ticket/122
and 
  https://trac.osgeo.org/grass/ticket/1452  [*]

pre-selecting datum transform terms on our behalf, while at the
same time removing all indication of what the +datum originally
was, really fouls things up for GRASS, and more locally for all
us EPSG file users in New Zealand, where tectonic shear from
the west coast to the east is on the order of about 4cm/year
over a distance of hundred km or so & the choice in the matter
has a rather big impact on our results / it's good to be forced
to specify which flavor your dealing with.

Ideally, for datums known to the osgeo stack I'd beg for the
+datum term be reinstated and that datum transform terms not be
decided on our behalf, with no way to back out of the pre-
selected choice.

More generally it means we can't really use the epsg file as a
semi-authoritative source of what some given epsg code refers
to anymore. :-(


--
ignoring my argument that it isn't really possible/a good idea,
and just assuming that it is possible to choose between 7-term
vs. 3-term vs. grid file as the sensible default, it is my
understanding that the 7-term solutions are in general only
valid (respected?) in a much smaller geographic area than the
3-term ones. aka 7-term does not necessarily mean more correct,
and if the epsg code covers a wide area, in much of the region
a locale-specific 7-term solution would be the worse choice
than a less precise 3-term solution. ISTR that Cliff(??) had
published an article about this paradox some time ago.
 


thanks,
Hamish


*see also
 https://trac.osgeo.org/grass/browser/grass/trunk/lib/proj/datum.c#L59