# Geometric locality as a mean to latency reduction in pull based protocols In this [post](https://ethresear.ch/t/selecting-optimal-outbound-neighbors-soon-for-fast-bandwidth-efficient-propagation-in-p2p-networks/23358) we discussed and illustrated how forwarding to peers with low RTT can lead to an overall reduction of latency (at constant bandwidth) in push based broadcasting. Some of the networking applications, such as the sparse blobpool design or [cell-level dissemination](https://ethresear.ch/t/gossipsubs-partial-messages-extension-and-cell-level-dissemination/23017) are however, pull based. This begs the question: *Can the insights we gained in our analysis of push based broadcasting be used to optimize even some of these pull based applications?* We believe that the answer is affirmative. The fundamental principle on which SOON is based is to propagate messages along paths with low stretch (see Figure 1 in the post and the discussion surrounding it). The same principle can be applied to any pull based protocol. Usually, such protocols include the exchange of information such that pulling can be effective. To see this consider two nodes $A,B$. In a symmetric protocol $A$ informs $B$ (and visa versa) on the data it wants and on the data it has. Based on this information $A$ and/or $B$ can then selectively provide actually useful data to their corresponding peer. It is clear that the exchange of messages informing peers about the data a node wants and the data it has is still happening on the basis of a push. Thus, the choice of which peers to contact can still be done on the basis of the principle underlying SOON. Further and more importantly, if a node selects the peers from which it is to pull on the basis of SOON, this also ensures that the forward path of the data (from the source to the node) uses the same low-stretch paths. Summarizing, we would say that the underlying principle governing SOON, i.e. propagating along low RTT paths can be exploited also in pull-based protocols which might become more relevant in the near future.