OPeNDAP plug-in issues

By enabling opendap_model in the compass_viewer, I can build HDF-compass on Linux. So I can test the opendap plug-in at least from my development environment.

In general,
One can open a URL served via DAP via "Open Resource". A simple netCDF style DAP service can be opened with the following limitations.
1) The current opendap model misses group attributes (at least for root)

The attributes under the root group(or global attributes) can not be displayed. It seems the feature is missing.

"
File "/mnt/scr1/kent/hdf-compass-kent/hdf-compass/build/HDFCompass/out00-PYZ.pyz/compass_viewer.frame", line 388, in on_menu_reopen
File "/mnt/scr1/kent/hdf-compass-kent/hdf-compass/build/HDFCompass/out00-PYZ.pyz/opendap_model", line 272, in **init**
AttributeError: 'NoneType' object has no attribute 'attributes'

Note: Pydap client on the command-line can successfully retrieve the global attribute. So this is an issue related to the opendap model.
Pydap output is as follows:

> > > dataset = open_url('http://test.opendap.org/dap/data/nc/coads_climatology.nc')
> > > print dataset.attributes
> > > {'NC_GLOBAL': {'history': 'FERRET V4.30 (debug/no GUI) 15-Aug-96'}, 'DODS_EXTRA'
> > > : {'Unlimited_Dimension': 'TIME'}}

I think it is promising to fix this issue. It seems the opendap_model just forgets considering the global attribute case.

2) Can not serve any object names that contain special characters

The opendap plug-in can only address the object names that include underscore,letter and numeric values(CF naming conventions). So essentially all HDF5 data served by the default option of the Hyrax OPeNDAP services can not be served since a "/" is always included in an HDF5 object name.
Symptom: the plug-in just opens the file without displaying any objects. The file name displayed by the plug-in looks like to be escaped with %...h5 for the dot in an HDF5 file name.

Pydap command-line client doesn't have an issue to open an HDF5 file served by the default HDF5 handler although I may see %.. displayed for the special characters inside the variable names.

3) Can and potentially will not serve compound datatype objects well because of the limitation of pydap

My version of the pydap(installed by using pip) can only handle a scalar compound datatype HDF5 dataset. So we should NOT expect the opendap module to be able to serve compound datatype objects well even if bugs are fixed.