HDF5

[Ray Osborn]

Some of you will have received the latest HDF newsletter which announces a
preliminary release of HDF5 <http://hdf.ncsa.uiuc.edu/nra/BigHDF/>.  I've
had a quick glance at it.  It's really a major revision which will require
a complete rewrite of the NeXus API when it's finally released.  The
impression I got is that it should make things a lot easier by simplifying
and harmonizing all the current data models into one unified object model.
It may even solve a number of our problems e.g. dimension scales are not
explicitly defined, but instead, you can create composite objects that
perform the same functionality, but without the name limitations.
Attributes don't even exist explicitly, but can be mimicked by attaching
data objects to each other.  Also, HDF5 might solve some of the storage
inefficiencies of the current SDS implementation.

My point in writing is that, as a potentially large-scale user of HDF, we
should get in our comments to the HDF5 development team as soon as possible
to make sure our needs are met.  Does anyone have the time to have a
critical look, and even to attempt to program in the new interface?  If
HDF5 is really coming soon, can we avoid some of the headaches of making
the NeXus API backward compatible e.g. by providing a simple conversion
utility?

Any thoughts?

Ray

------------------------------------------------------
Dr Ray Osborn                  Tel: +1 (630) 252-9011
Materials Science Division     Fax: +1 (630) 252-7777
Argonne National Laboratory    E-mail: ROsborn at anl.gov
Argonne, IL 60439-4845