[MetaCRS] Coordinate offset after transformation from 4267 to 4326

[Martin Desruisseaux]

Le 26/05/16 à 14:56, Finn, Michael a écrit :

> Thank you for this detailed explanation.

My pleasure :-)


Le 26/05/16 à 15:12, Jody Garnett a écrit :

> I would still like to learn if there is a way to check what PROJ4 is
> up to, as this is the first time I have worked with grid shift files I
> have no confidence I am doing so correctly.

This would indeed be a useful information. In my understanding (I may be
wrong), PROJ4 is what EPSG calls an "early-binding implementation",
meaning that it relies on the information attached to the CRS like the
TOWGS84 element except maybe in a few special cases. Early binding
implementations are often designed on the assumption that some CRS
(typically WGS84) can be used as a "hub" or a "pivot CRS" for all
coordinate transformations. Whether such assumption is okay or not
depends on the desired accuracy.

By contrast, a "late-binding implementation" (in EPSG terminology)
searches information about coordinate operations in the EPSG database
only when the full (sourceCRS, targetCRS,
other-aspects-like-area-of-interest) tupple is known. So the
transformation from A to B is not necessarily the same than the
transformation from A to WGS84 to B. GeoTools and Apache SIS are
late-binding implementations.

EPSG recommends late-binding implementations, but those implementations
are not easy to get right. In any cases, PROJ4 and GeoTools
implementations are different enough that in order to perform comparison
between them, we need to know the transformation paths (operation method
and parameters) that both of them selected.

    Martin