tinco
@tinco thanks for the api. I just started digging into it. Questions/feedback:
Is there a way to tell participation level (watcher/archiver/basic/full)? https://www.stellar.org/developers/stellar-core/software/admin.html#level-of-participation-to-the-network
Looks like you get org info from known validators list. This seems useful for informational purposes in exploring but not reliable for calculating trust or resilience as it's easy to have many nodes not listed or listed under different info which would seem unassociated.
Is accepts_connections just the most recent value? Looks like some nodes where that's false have true in history? If so seems more useful to show an average and then the full history.
trust_table is interesting. Looks like = 1.0/threshold w/o inner quorumset and =1.0/threshold/innerQS-threshold if the node is in an inner quorum set. There might be a bug enhance_accounts line 29 see the last 3 table values for validator GCKWUQGSVO45ZV3QK7POYL7HMFWDKWJVMFVEGUJKCAEVUITUCTQWFSM6
Imagine a case of a validator with qs like:
root-qs (threshold 1 of 2)
inner-qs A (threshold 2/2: nodeC, nodeD), inner-qs B (threshold 1/1 nodeD)
This is equivalent to quorumset of 1/1 of nodeD
AKA
( C && D ) || ( D )
simplifies to
D
but in your forumula you're giving trust to C. Although in my analysis no validator actually reuses a node in its quorum set.
Also, should trust be the same for a node that is in qs (threshold 1 of 1 nodes) vs qs (theshold 1 of 10 nodes).
This is the reason I came up with with:
https://github.com/foundero/NodeStar#math
- trusing_nodes - this is what I'm calling direct incoming. Also consider indirect incoming where you count all validators that indirectly trust this validator by following each quorumset node's quorumset. I think this get more towards something like pagerank. I also calc indirect outgoing... these show emergent properties of the network which I'm calling clusters (kinda like tiers) where each represents a set of validators with identical indirect incoming+outgoing sets. Interestingly there are not very many clusters, so it makes the entire network a bit easier to visualize: https://nodestar.info/#/clusters/1
hit me up on keybase/slack if you want to discuss