dev

I am studying the paper "Swam: a family of access methods for similarity-search in peer-to-peer data networks", and got the following questions:
1. is the lp-norm distance computation-efficient? what's the difference between Euclidean and it on the computation cost
2. based on small-world theory, an O(logN) is acquired, but I still don't know the proof detail, is it a breakthrough over Jon Kleinberg's distributed algorithm?
3. since there is no assumption on the data distribution, how to establish long-range links is interesting, again, it's related with their algorithm on the modified small-world theory, I think
4. a related question is whether a point lattice is a voronoi graph, because this paper employs voronoi graph, and I want to induce some thoughts from regular grids, is it possible that there is still a trap, I need to study voronoi graph more
5. turn to our project, it seems that the following problems are interesting on their own part, first, the degree of each node, second, the dynamic dimensionality, third, the evaluation metric, all these should be considered carefully

today I am also reading Viceroy, a butterfly network based routing protocol