Thanks for your reply. As I said, if I do it from the scratch, I’d likely be biased towards their solution. It feels somehow awkward to look for a different way of implementation only to avoid legal issues. In this part (LU decomposition), their implementation is not very original.
In the meantime I found in the library the book that they cite at the end of S.9, by Dennis and Schnabel. The linear search algorithm in NR comes mostly from there (I still need to compare the derivation, but it seems to be the same thing). So, the question is really how much one can claim rights on implementation of well known methods described in literature?
Thanks for mentioning minpack. I’ll check carefully what is in there.