User interface, client/server issues, and FEFF

[Bruce Ravel <ravel@phys.washington.edu>]

Howdy,

Lately I have been thinking a lot about FEFF interface issues.  This
is a big topic.  That it is a big topic should be clear by the large
number of text lines that have been slung back and forth between us in
recent weeks ;-)

Like any complicated issue, this one has many facets.  I have been
thinking about what is needed by the FEFF user -- both the novice and
the power user.  I am never very content thinking only about what the
novice needs because a novice-level application will probably not be
very useful to me.  Thus I like to think about the power user, but I
try to keep an eye on the needs of the novice.

To this end I have been shaping a client/server model in my mind.  The
basic idea is to have a separation between the program that the user
uses to prepare for a FEFF calculation and the program that
actually performs the calculation.  This is certainly desirable in
that it relieves the API to FEFF's functionality from having to serve
the needs of user-friendliness.  But the appeal of a client/server
model becomes huge when the recent development of message passing
parallelism is considered.

Do you have a 64 node cluster on your desk?  I sure don't.  But I
would certainly like to be able to compute on one.  It would sure be
nice to use a GUI on my computer to prepare the parameters for a
calculation or a series of calculations and use a big computational
monster at some other location to actually do the heavy lifting.  That
screams client/server.

I have written a series of documents -- I suppose you can call them
whitepapers -- and posted them on my web site.  These whitepapers
summarize my thoughts on how a client/server interface to FEFF might
be organized.  Here's the URL:

   http://leonardo.phys.washington.edu/~ravel/misc/feff_interface/

There is no code there.  It is just a bunch of pages of me describing
ideas.  It is kind of a big brain dump.  Hopefully this will prompt
discussion or even encourage someone to start working on this with me.

I'm going to be out of town for two weeks starting tomorrow.  So if I
don't answer your criticisms soon it is not because I'll be ignoring them
but because I will not have yet have read them ;-)

Regards,
B

-- 
 Bruce Ravel  ------------------------------  ravel@phys.washington.edu
 U.S. Naval Research Laboratory, Code 6342      phone: (1) 202 767 5947
 Washington DC 20375, USA                         fax: (1) 202 767 1697
 WWW: http://feff.phys.washington.edu/~ravel --------------------------