From sci.math.num-analysis Thu Mar 18 08:41:34 1993 From: fichtner@acs.ucalgary.ca (Horst Fichtner) Subject: 3-D PDE on non-rectangular domain Organization: The University of Calgary, Alberta I want to solve numerically an elliptical PDE on a three-dimensional domain.Rest of message deleted. The request was answered on 93 Mar 19, as follows:
From: Han.deBruijn@rc.tudelft.nl (Han de Bruijn) I am in the process to develop such a code. If you describe to me in some more detail what your problem is, [ ... ]The collaboration then started by "setting the stage": how to explain the physics of CR to a non-astronomer in the first place? It also became clear that the CR problem could not be solved by a kind of "general" code. Hence it was decided to develop a quite dedicated computer-program instead. The exchange of mathematics "at a distance" was accomplished in an efficient manner with help of \LaTeX. The following splitting of responsibilities has proven to be effective. A numerical method for solving convection-diffusion equations of the CR type can be developed, quite independently of whatever physical details are actually involved. Because the physics is mainly in the coefficients of the equation, the latter can be separated a great deal from the equation itself and separately coded. Modular programming is not luxury, but merely a necessity when people with such different backgrounds are going to collaborate! At last the program, with exception of the "coeffs" routine, was "shipped", by e-mail of course, from Delft University to the University of Maryland in the beginning of May 1995:
From horst@aix.umd.edu Sat May 6 21:54:25 1995 Subject: Nice performance...This concerned the version with the direct solver in it. The iterative programs were sent a couple of months later. And the following response was received:
From hfichtne@wam.umd.edu Fri Oct 6 22:10:44 1995 Subject: Flabbergasted...