In the paper, we chose seed-vertices such that (i) the resulting Steiner-tree is a connected-component, (ii) majority of seed-vertices are not directly connected in which case shortest-path computation could converge faster (to avoid a potential favorable scenario). Here, we discuss two alternative seed selection approaches that we have explored: 1. Uniform random selection of seed-vertices from the largest connected-component. This does not always respect requirement (ii) above, especially for smaller graphs e.g., CiteSeer, MiCo.