I decided, I want at least a day to relax a bit, meaning work on kmath a little bit more and I have a design question. If you remember, I've introduced a matrix feature construct, meaning that each matrix has additional metadata collection with optional features describing the matrix (sparse, diagonal, singular, etc). It makes perfect sense for matrix, but it is possible to add it for linear structures as well. The question is: does it make sense? It is possible to use those features to store some meta-information like structure mean, min max or some kind of more sophisticated info.