posting one image at a time just in case, here is the original diagram I came up with as an idea. The goal is to find an algorithm to reduce the leaf nodes in a graph when a new node is inserted at an arbitrary position in that graph.
#Peer-based algorithm for flood graph fill
weary bridge
is my current code - which produces this:
a significantly more tree-like structure: which is undesirable for many reasons: one being this -
nodes like peer-15 in the picture here: are isolated island connections in the graph, if they are removed, suddenly you have two graphs: which is a failure state.
My general strategy is this so far:
when a new node is inserted (somewhere in the graph), a node takes it and processes it, to either add to the graph connected to itself (and maybe a piggyback connection), or forward the node with some found connection(to piggyback) to a connected node in a router.
each node has two 'routers' A, and B. which query a short list (demo is 3), to go to the smallest score (a score is how many connections does it's{the processing node} connected nodes have) or in a round robin if all scores are equal.
The thinking behind this is if the connected nodes total connections is smaller - it will always fall towards the leaf node that is nearest, filling the gaps quickly... that is not what is happening. Instead it makes 'roads' and branches into trees...
The idea of having a bi-router system was to organize the network as the original diagram showed above - but it may be possible (better?) using only one type of connection and not organizing them into polar pairs... but polarity seems like it would be a useful property...