[PROJ] NTv1/NTv2/CTable2 and west-oriented longitude shifts

[Even Rouault]

Hi,

Another messiness of NTv2 I (re)discover. As NTv2 was designed for Canada, it 
uses a "west longitude is positive" convention.
That applies to the extent of the grid, the order of the columns... and the 
longitude shift values.
NTv1 also uses that convention.

Our venerable CTable2 format uses a more classic convention of negative values 
for west longitudes and use a west-to-east column ordering, but has also 
inherited that annoying characteristics that the delta longitudes are postive 
to the west... So this affects the +proj=deformation . Kristian, are you aware 
of that ? But as you hijack CTable2 to store values in mm/yr, you probably 
compensated that issue with the negative sign.

So what about GeoTIFF storage for horizontal grids ? Do we want to perpetuate 
that tradition of positive longitude shift meaning correction to the west ? Or 
use the "natural" convention: positive longitude offset values are corrections 
to the east ?
It seems that at the very least I should add a new metadata item: 
positive_value=west/east to document the convention.

Perhaps for grids converted from NTv2, I will keep the original value and set
positive_value=west so that it is obvious we didn't alter the values.
People wanting to use the "natural" convention will be strongly encouraged to 
use positive_value=east (which will be the default)

Thoughts ?

Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com