It appears that there are quite a few nodes that are out of sync, which makes sense based on the longer block times announcement. I noticed since dynamic issuance was turned on (block 718959) the bandwidth on my nodes are up considerably. It appears the offenders are the nodes that are out of date (sorted on telemetry by uptime) and everyone one of them are downloading incredible amounts of data (most at hundreds of mbps).
Would it make sense to drop connections from older incompatible nodes (or incompatible runtimes) going forward to protect users from excessive bandwidth and connections from nodes that may not be able to sync?
It is not as easy to do, one of the things on live distributed network is that you can’t force people to upgrade.
However since those who don’t upgrade miss rewards all this time and the rest earn more (proportionally) in the meantime, I think they will upgrade soon and situation should improve. I’d rather not introduce breaking networking change just to address this.
Do nodes see the version of connected peers? If so, having the ability to drop connections from problematic nodes if the need arises may be advantageous, and could be done from the client itself.
I was just thinking, could something similar be used as a potential attack vector by intentionally introducing problematic nodes to the network?
Yes, sometimes the upstream of my Node server can also be fully occupied. Normally, only a few hundred KBs are needed, so I think this is a risk unless we apply rate limiting to Node.