[buescher]

There was a numerical analyst at JPL back in the day that absolutely savaged NR. Remember, the kind of engineering they do at JPL has to be correct, not just plausible. Also, noncommercial alternatives to things like Matlab (octave, the python interfaces to Netlib libraries, etc) weren't as mainstream then, and science and engineering research had a bigger "get a book and roll your own" ethic. https://web.archive.org/web/19990202153232/http://math.jpl.n...

I've seen it breed horrors - "senior" people instructing interns to implement its matrix inversion routine line-by-line in Matlab. Today it would be Python.

Oh, and here's a list of alternatives, from the heyday of NR: https://web.archive.org/web/19990202141044/http://math.jpl.n...

[B1FF_PSUVM]

I just like this phrase from the Usenet comments in the first link:

> Their code has the implicit assumption that all elliptic problems have homogeneous Dirichlet boundary conditions!

[FabHK]

I think the basic problem is that the algorithms were for exposition, not production quality (they would not guard against under/overflow, edge cases, etc.)