3 ms·
> P2P does not scale to large numbers of participants I'm curious about this statement. Intuitively it seems there is already an algorithm to near-optimally di
by imglorp 7y ago
> P2P does not scale to large numbers of participants
I'm curious about this statement. Intuitively it seems there is already an algorithm to near-optimally distribute updates to a sparsely connected mesh network. You'd want two or three connections from each peer to another, not N:N. Optimal is not necessary as long as you have at-least-once semantics on message delivery; a node can just ignore messages it's seen already, as long as it's not too often. Is there a name for this existing algorithm?
On top of that you could overlay a symmetric session key and you'd have full E2E. Then central server would be session initiation only--which could still do authn/z to hand out session keys--and billing/metrics.