[Xrays@aps.anl.gov] 2006 XSD Scientific Software Workshop User Survey

[Matt Newville]

I see two main areas that the APS could focus on to really help the Users need
for scientific software:

First, the APS could help develop and encourage use of a common set of tools
for staff and Users to write software in.  I wouldn't exactly call it
a Framework,
but that might be close to the idea.  The primary needs here are visualization
and data handling, and so probably ARE closely related to data
collection.   I think
the APS should not try to dictate a "Total Solution" by itself, but
provide basic tools
(and support and documentation for them) and to foster the development of
a community for x-ray software.

Personally, I think the APS should encourage and promote the use of Python and
C for this work, but that bridges to other languages must be available
(or fairly
easy for someone else to make).  That is, I think Java (or .NET) would be poor
choices for getting community involvement from User scientists.   I
think we should
be looking beyond the x-ray ans neutron communities as well (astronomy and
computational chemistry, bio-informatics, and geographic information
systems come
to mind), as many of the tools needed are fairly common to many
scientific fields.
As a first priority, I think that embracing the challenge of a good
data visualization
package for 1-D, 2-D, and image data would be the right goal.

Reasonable starting points exist and are actively developed and
supported in Python:
  http://scipy.org/
  http://matplotlib.sourceforge.net/


Second, and completely unrelated to the discussion above, the APS could
support a "User Facility" level interface into modern computational
chemistry and
physics codes,  such as for DFT calculations.  Much of the work done at the APS
could greatly benefit from having a computational component.

Several groups are doing such work now,  but the barrier to these DFT and other
codes can be pretty high.  I'd guess that most people including such
calculations
in their work are collaborating with computational chemists.   It
would be easy to say
this is how it should be, but just as the APS provides support labs to
users for sample preparation and some detectors and other equipment
for general use, it might be
worth considering having a supported User Computational center for such high
performance computations.


--Matt Newville <newville at cars.uchicago.edu>