I noticed that, but he's doing a single REST query for every timing: https://github.com/jhb/neo4j-experiements/blob/master/query_friends_neo4j_rest.py#L24 http

I noticed that, but he's doing a single REST query for every timing: https://github.com/jhb/neo4j-experiements/blob/master/query_friends_neo4j_rest.py#L24 https://github.com/jhb/neo4j-experiements/blob/master/query_... There's no way that a single HTTP request is responsible for a difference in hundreds of seconds (his benchmark code isn't really sound, but I don't see how this can affect a difference so striking)