I have some basic question, that I kind of know the answer to but need clarification on:
1- What percentage of the stakes must be online for the network to keep generating blocks(for example in a local network).
2- Does the gossip protocol in Algorand offers some kind of advantage over regular TCP protocol P2P in propagation delay or other features?
1 Like
I’ll try to answer your questions -
1 - 80% or more of the online stake needs to be online for a smooth, uninterruptible generation of blocks. The total amount of the available stake doesn’t really matter. What matter is the amount of online stake, and what percentage of that stake is actively voting. An online account that doesn’t vote is considered by the protocol as malicious ( as per the white paper ).
2 - No. The Algorand gossiping protocol is implemented using websockets over HTTP ( TCP ). The reason for that is that TCP is supported everywhere, and all the existing network proxies / routers already there to support it. Note that the network library does make a “good use” of the given WebSocket channel, in the sense that it knows to prioritized queued messages, etc. Using UDP for the messaging could work for some cases, but fail on others : for example, adding a proxy for a UDP based connection might not be such a trivial task.
1 Like