Hi @timk11 thanks. To clarify, decentralization coefficients are currently calculated by country, node provider, data center, and data center provider. So swapping a node machine from, for example, Colombia to Germany does not improve or reduce the decentralization coefficients if there is not already a node machine in the subnet in Germany. Most of the time, the node replacement proposals are for two reasons:
- a node is degraded or dead, so it is replaced by another node from another (or the same) node provider, allowing the node provider to redeploy the node.
- a node is removed in order to allow regular maintenance.
I noticed as well that both proposal 131704 and proposal 131705, are replacing a node that is currently downgraded or dead. The other node in 131704 (in Estonia) and in 131705 (in Florida) currently show healthy. The reasons for swapping a second node machine is that in the currently implementation of the DRE tooling, node machines are always swapped in a pair of 2 nodes.
The reason for this is that with a subnet of 13 node machines, it is always safe to replace 2 nodes without compromising the subnet, while at the same time in almost all cases - when there is only one unhealthy node to replace - a second node can be found that improves the decentralization. But in the case of 131704 and 131705, it could not found a further optimization; in that particular situation, the second node to replaced is chosen at random from available nodes, that at least does not decentralization coefficients. In the case of 131704 and 131705 these were nodes in Switzerland and Germany, which at least did not make decentralization coefficients worse. In the case of 131703 as another example, the second node chosen was the same as the node that was already in the subnet, so no second node machine was swapped.
Agree that this is not perfect, and the tooling could be further improved, amongst others to take into account geographical areas such as “EU”. This might be a topic that needs to be discussed further, and whether additional decentralization coefficients needs to be added upon and implemented.
Furthermore, the summary text in the proposal that is automatically generated and says “improving decentralization”, should actually be “optimizing decentralization”, because sometimes no improvement can be made.
As for the two proposals, these were rejected by the Foundation, but for another reason: the nodes added nodes from a data center in Munich (MU1) that is being decomissioned, as discussed above.
Thanks again for reviewing of these proposals, it will definitely help improving the quality of the proposals and the decentralization of review of these proposals. At the same time we will be working on improving the auditability of each of these proposals.