Author: Gregory Maxwell
Published on: 2011-07-01T16:23:59+00:00
In an email conversation, Christian Decker suggested that multithreading the RPC and adding Keep-Alive to the code would be beneficial. However, these changes require aggressive testing to ensure they don't expose race conditions in the code. The multithreaded RPC patches have only been run by miners so far. The proposal for a 'hub mode' to optimize connection logic for all nodes was not accepted by everyone. The concern is that it will consume more resources than it adds to the network. Instead, miners should manually addnode each other to improve block propagation speed. Addnode logic should be fixed to reserve connection slots for addnoded nodes and try to keep connecting to them. There is a lot of room for longer-term improvements to the connection and forwarding logic, but there isn't a good way to test and validate changes.
Updated on: 2023-05-17T18:13:13.020025+00:00