Time to get science majors cracking or at least hacking.

Mathematica is too obtuse, MathCAD is too restricting for anything beyond the simplest few-line applications, believe me I've tried both in Astr 137 and Astr 142.

From my experiences in Astr 137, I've developed the following Minimialist Approach to Scientific Computing.

• Pick a language that is high-level, strongly typed, extensible within reason, has a small number of keywords, and supports Numerical Recipes. Java is too extensible, C# is too Windows specific, C++ has too many operators and like Java is too extensible. But Fortran 95 in its streamlined and free F incarnation or full GNU g95 compiler) fits the bill nicely. In particular, Fortran 95's powerful array language leads to very transparent image/spectrum processing programs that are as high-level as e.g. IDL. No loops! Not much boilerplate! forall and where, masks, lots of intrinsic functions, support for Numerical Recipes in Fortran 90 that generally get the job done. (Of course, someone has already typed in all the useful NR routines), and g95 now includes the proposed Fortran 2008 coarray extension for transparent multi-processor/parallel computing.
• Execute in the dismal but familiar Windows XP environment so you don't intrude on the generic CCI laptop load. Use MinGW and MinSys, available here.
• Use a high-level graphics library with high-quality output from a very small number of simple function calls. DISLIN works well for Windows and Linux under Fortran 95, has an easy to program widget set for text input, and works effectively with cgi-scripts such as this one. So students can generate dynamic gif/jpeg/png graphics on their Web pages with no effort at all.
• Give them a very high-level image display function, a couple of functions to hide all the complexity of the cfitsio (use v5.03 with MinGW) library (they only need my simple rfits routine included here for image/spectrum/datacube reading/writing.
• Add the SciTE syntax-aware editor to handle compile/link/execute.
• Bundle all Fortran 95, FITS file access, and graphics code into a single module that they simply USE in each program,
• I'm still evaluating debuggers, but MinGW has several candidates.

If everyone adopted this approach, along with MathCAD for the simple stuff and LabVIEW for lab work, our Physics majors would be developing very sophisticated applications and would benefit from a growing set of our pedagogical programs. How about it folks?