Forget counts given by vault-count.safenet given a few days ago. This program crawls the nodes by recursively getting the nodes in close groups. The problem is the initial set of nodes I start from:
-
a first version of the program was starting with nodes in my own routing table. The results were OK in a little local network of about 20 nodes. But this algorithm misses a lot of nodes in a bigger network.
-
a second version was starting with close groups of about 30 random addresses. The results were much better but were fluctuating
-
the latest version starts from 32 uniformly distributed addresses, which gives stable results and I am confident that there are really currently about 140 nodes in the network.
All this is about line labelled Current total nodes.
For the line labelled Locally estimated number of vaults I do an extrapolation from the distance of furthest node in my close group as explained here, except that I don’t subract a fraction of standard deviation. This value is highly variable and should be averaged on several vaults but I have only one.