Before we kill pkg-science, are people ok with pkg-scicomp including
e.g. graph theory? I.e. stuff outside the normal "numerical" idea
that some people have about scientific computing? I certainly don't
mind, but when I hear "scientific computing", I think of numerics.
I guess you are referring to metis and scotch.
These software are actually quite important in numerical computations for example in mesh partitioning or linear algebra.
Software/libraries like petsc, trilinos, suitesparse for example are using them and they are all about numerics.
Best regards
C.