[FWTools] OpeNDAP and GDAL - I am going brain dead

[Frank Warmerdam]

Michael Sumner wrote:
> Do you have the OpenDAP driver support built?  Does gdal_translate 
> --formats report OpenDAP as supported? If so you are much farther ahead 
> than we ever got.
> 
> When a (much more experienced) friend of mine looked at this with me, it 
> came down to incredibly difficult-to-debug problems involving OpenDAP's 
> dependencies when compiling from source. This was using Cygwin: his 
> conclusion was "wait until someone fixes xyz".  I think "xyz" might have 
> been libcurl, or something that depends on in turn.
> 
> We didn't get close to compiling the OpenDAP sources - which is the 
> starting point you would work from in order to build GDAL support for 
> that driver.
> 
> I'd rather lobby for worldwide, permanent abandonment of this format, 
> than to try to figure it all out. It's incredibly limited and painful to 
> use even when it "works", requiring endless augmentation to obtain 
> usable data, in my experience (which was by running Matlab processes on 
> Linux from R).

Mike,

While I sympathize with your pain, and I've also had my share of
frustrations (some due to limitations in my GDAL side DODS support
code) I think that OPeNDAP is a very sweet protocol for accessing
science data over the net.

> But if you make any progress I'd love to hear about it.

The FWTools build for linux should have working DODS support.  For Roy's
server I suggested the following URL to pass to GDAL.

http://oceanwatch.pfeg.noaa.gov:8081/thredds/dodsC/satellite/AG/ssta/mday?AGssta[3][0][-y][x]

One item that wasn't immediately obvious to me was the requirement for the '?'
in the string.  I first tried the same url, but with that as a '.'.  This
might work with DODS just fine, but unfortunately GDAL needs the '?'
as a clue where to cut the "projection" part of the string off so
it can parse out the [y] and [x] portions.   If you don't do this you
do *not* get an obvious error (unfortunately).

The other issue is that for now you have to fix all dimensions other
than x and y, so in the above case we take a specific time slice [3]
rather than treating all time slices as bands as might be desirable.

I think a "hints and tips" topic in the GDAL Trac wiki for OPeNDAP/DODS
might be a good idea.

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | President OSGeo, http://osgeo.org