Due to our own lack of experience/hardware/software, we definitely need
volunteers (we'd be willing to help where we could of course) for this stuff:
o some sort of Windows interface - preferably velvet for Windows, but
it could also be xfelt for Windows to start with (that would be
a lot easier).
o as part of the above or as a separate undertaking, we need someone
to compile the command line applications under DOS such that they
will run on a 286 and while Windows is running. The current binaries
are built with DJGPP (gcc for DOS); the kind of binaries I'm talking
about would need to be compiled with Microsoft or Borland C I think.
If someone wanted to write pre- or post-processing applications for
DOS that would be cool too. Post-processing would be particularly
easy if someone just wanted to parse out the standard felt ASCII
output. You could start with feltvu as a simple starting point.
It might be nice if someone could port feltvu to a different
compiler as well. I think most major DOS compilers probably
handle graphics drivers in a more reliable way than the current
set of DJGPP graphics lib drivers which people have had a lot of
o all of the above kinds of things (particularly just binaries of the
command-line applications) for the Mac.
Things that maybe we'll get to someday, but if you want to try them first
then please feel free. In no particular order:
o make use of the Layout widget for geometry management of the main GUI
o more intelligent use of the command line ... command completion
and argument specification ... type node hit space,
get a "list of options" type delete hit space, get a prompt
for a number, type number hit return, etc. In other
words make it something along the lines of some of the
popular CAD systems.
o improve plotting of displaced shapes. This probably means that
element writers need to provide a pointer to a function which
knows about the shape functions for an element and can fill in
a vector over the area or length of the element based on
nodal displacements which, when plotted, will reflect the true
displaced shape of the element ...
o add snap to tool capability
o robust quadrilateral mesh generation, 3-d mesh generation, etc, etc.
o additional higher-order elements
o fold Timoshenko beam theory into the standard beam, i.e., if
kappa or nu is given then expect a G and use Timoshenko
theory instead of Euler-Bernoulli.
o put some real effort into internationalization. The current
functionality is there because it was easy. More robust
implementation would mean maintaining a table, in code probably,
that mapped all possible input and as much output as possible
to a language determined by the user.
o plug in something more universal than PPM to replace GIF - this will
have to wait until some sort of agreeable format emerges of course.