> As a personal anecdote: I was reading a good book about numerical > computation the other day. Everything was nicely derived and all that. But > when it came to the C library that the respected professor had developed, I > couldn't but think: this is absolutely terrifying. I mean, the code examples > violated practically every idiom of C; several functions took well over > twenty parameters, numerous libc-functions were severely misused, the type > system was a mess, and so on. > > - Jukka.
Do you remember which book? You've gotten me interested in which aspects of particular software packages you think are terrible. I'd like to take a look at them to see if I can learn what *not* to do. @Kris: What is it about Gnuplot that's generally awful? Plotting and graphing seem like they'd be less complex than operations like, say, factorization, integration, or solving systems of DEs; do you think it would be possible to write a less-sucking plotting package? Any thoughts on what such an undertaking would involve? I ask because you mentioned CASs being in a class of software which might not take well to simplification. Best, A.J.
