Hi Ethan,
I see that netcdf-java client is designed to avoid downloading the
entire data onto client. Otherwise, in case of a simple fileServer, the
client could download the entire data before decompressing it (if it is
compressed).
Upendra
Ethan Davis wrote:
Hi Upendra,
On 6/8/2010 12:11 PM, Upendra Dadi wrote:
Hi Ethan,
It does help. Thank you very much. Just to confirm if my understanding
is correct- the client software deduces whether the URL is an OPeNDAP
URL or not by making additional requests to the server and if it is, it
requests the data in uncompressed format from OPeNDAP server which the
OPeNDAP server would be able to handle whereas other servers wouldn't.
Yes. OPeNDAP is a data access protocol that allows clients to request
descriptions of the dataset and subsets of the actual data values (more
at www.opendap.org). The server must therefore be able to read,
understand, and extract the data from the file.
Access through the "fileServer" URLs are simple file transfer over HTTP.
So the server doesn't have to understand the data file it is serving.
(The netCDF-java library actually uses HTTP byte range requests to not
get the entire data file. But the server doesn't know anything about the
file except for the bytes it contains.)
Ethan
Thanks,
Upendra
Ethan Davis wrote:
Hi Upendra,
Are the data files accessible from foo.com through OPeNDAP? If so, try
using the OPeNDAP URLs in the NcML rather than the file server URLs.
With OPeNDAP, whether the source data file is compressed or not
shouldn't matter as that is all handled on the server.
When accessed through the file server, the client software is reading
the bytes from the compressed file over HTTP. When the netCDF-Java
library tries to open a non-OPeNDAP HTTP URL, it assumes the resource is
an uncompressed file (for efficiency and other reasons). When it tries
this on a compressed file, it doesn't recognize the first few bytes of
the file as the netCDF header and so fails to open the dataset.
Hope that helps,
Ethan
On 6/7/2010 4:05 PM, Upendra Dadi wrote:
Hi,
I am trying to aggregate some netcdf files thought ncml. I am getting
an error like this when aggregating:
Cant read http://foo.com/thredds/fileServer/examples/test.nc.bz2:
not a valid NetCDF file.
But I am able to open the same file on THREDDS server. Any suggestions
on what might be going wrong?
By the way, does THREDDS download the entire netcdf file before
aggregating? How do I control how long the temporary dataset stays on
the server?
On 6/8/2010 11:10 AM, Upendra Dadi wrote:
Hello,
Can someone please help me here with my ncml issue? My THREDDS server
is unable to read remote compressed netcdf files present in the ncml
files. However, it is able to recognize remote uncompressed files as
well as local compressed files. Does ncml working with remote compressed
files? Is there a TDS setting I should be looking at? Thanks!
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit: http://www.unidata.ucar.edu/mailing_lists/
