[PROJ] NOAA VDatum Tidal Grids and PROJ CDN

[Even Rouault]

Eric,

12 GB is a large volume compared to the current 500 MB of PROJ-data. I don't 
think we would want to distribute that in the .tar.gz/.zip archive we build 
from https://github.com/OSGeo/PROJ-data which contains the 500 MB of data, and 
is already seen as too fat for some users when offline use cases must be 
covered (e.g. QGIS standalone installer)
Incorporating them directly in that github repository would also be probably a 
bad idea, as it would make it unconvenient to manage (any contributor wanting 
to add a 1 MB grid would have to fetch a 12 GB repository...)

However, we could potentially import those grids on the AWS S3 storage behind 
cdn.proj.org (unless you have a cloud storage publicly accessible with those 
grids), and have records in the PROJ database binding the name of the grids as 
found in the EPSG database to the URL where they can be found (which is the 
business of the
https://github.com/OSGeo/PROJ/blob/master/data/sql/grid_alternatives.sql 
table). 
Having those datums, CRSs and grids registered in EPSG is indeed a prior step 
that would make it much easier to make them available to PROJ users, as 
normally the only remaining step is to do that grid name -> URL mapping. There 
might be some small code enhancements to do in PROJ source code depending on 
if a new EPSG transformation method is created for those grids.

Opinions of other PROJ stakeholders welcome on how to handle such dataset.

Even

> Hello,
> 
> I spent some time the past few days reading RFC 4 and some related threads
> to the PROJ CDN and PROJ-DATA repo.  I feel like I might be opening a can
> of worms here (in that this conversation might have already happened with
> someone else from NOAA), but using the CDN for distribution of our Tidal
> Datum grids seems like a useful step to take, both for our
> internal software development efforts and for other users who might want
> these grids.
> 
> Where does this group think these Tidal Datum grids should go?  Should they
> be included with PROJ-DATA?  The grids in question are ~12 GB in size,
> which might be the stopping point here, although with the dynamic CDN
> download option, maybe the total PROJ-DATA size is less important now.
> What would other users want us to do with these grids?  How could we
> provide this data to benefit the PROJ community?
> 
> In addition, we are in the process of compiling a metadata package to
> formally register these vertical CRS/datum transformations to EPSG, would
> using the CDN as a distribution mechanism + EPSG registration allow us to
> use these transformations in PROJ?
> 
> Welcome any advice on the topic.
> 
> Thanks,
> Eric


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