We have different security goals here, but I’ll start from liveness since you picked up the term Natamoto coefficient.
Yes, liveness could be violated by 1/3rd like all chains who require Byzantine agreement. It’s worse really: All blockchains remain vulnerable to simple denial-of-service (DoS) attacks, so really the true liveness level for all blockchains is zero validators, because none of us can build infrastructure like CloudFlare. Now major DoS attack could require burning a couple zeros days for some popular IoT devices, so wild guess this attack costs maybe $250k, and requires some networking knowledge and a well oiled botnet. Against proof-of-work chains you’ve much cheaper options, so zero feels excessive here.
Although Elves weakens liveness, we’ve actually taken liveness much more seriously than others overall: Our gossip topologies do balance these liveness considerations, even if not perfectly. We carefully limit the blast radius for cross chain messaging attacks. Alistair Stewart and Rob Habermeier tool liveness seriously in Grandpa. Rob Habermeier & others worked hard weaving “politeness” into our protocols. Aside, JAM has not done this yet, so that’ll add some time onto their deployment timelines, but they can see how Polkadot achieved politeness. We should’ve much stronger liveness than most of our peers.
Now soundness & safety are much more important properties than liveness, which includes events like double spending. As a rule, Byzantine agreement protocols have only 1/3rd margin here too, but they allow different 1/3rds in different places. This major improvement seems like the big reason why Ethereum switch away from proof-of-work, with a bullshit-ish 50% margin, to proof-of-stake with a more defensible 1/3rd margin.
We have safety & soundness against 1/3rd actively malicious, like other Byzantine agreement protocols, because ELVES costs us almost nothing there, unlike say EPFL’s OmniLedger. Aside: If we’re eventually successful enough to go beyond say 500-1000 parachains/services doable on one relay chain, then we’d likely weaken soundness to 80% of 1/3rds or 26% overall, in order to have multiple relay chains ala EPFL’s OmniLedger and ensure less than 1/3rd malicious lands on any one relay chain.
Anyways, there is a human factor behind all these numbers: Who are the people really? Why do we think they are distinct? Do we think they know what they’re doing?
In our case, the 1KV and DN programs have helped us diversify the validator operator pool, including by having affirmative action for validators in under represented jurisdictions. In fact, we lowered our raw DOT vulnerability threshold for years by running 1KV and DN, but this built up quite a few validator operations across different jurisdictions, so longer term this made us much stronger.
We’ve shut down 1KV and DN for the moment, because we’re reworking the staking notions. We’ll incorporate related diversification ideas into whatever nomination system emerges here. In fact, we’re doing the staking redesign because we no longer believed the nominators were doing their job correctly, aka nominators were not assessing competence, independence, and character.
Anyways all of our peers have exactly the same problem, but afaik few of them seem take these questions seriously like we do, not that we have perfect answers either, and ETH going PoS sure earns them points.
Also..
The term Nakamoto coefficient was designed to conflate liveness and safety, because Bitcoin stupidly conflates them, and Bitcoiners can more easily argue for liveness.
I suppose the Nakamoto coefficient of bitcoin has improved from one guy at each of 2 mining pools to one guy at each of 4 mining pools over the last several years, no? It’s mining pools you need for double spending, not miners.
In fact, Bitcoin has stupidly kept liveness and soundness both at claimed 50%, but not requiring Byzantine agreement. Yet, really various attacks that drop bitcoin into the low 20%s, which gives them a Nakamoto coefficient of maybe 1 even for soundness, aka double spending. lol
I’ll refrain from criticising ETH too much here, but our nugget BLS paper was exactly a criticism of how BLS signatures forced some centralisation upon them.
Anyways..
We could easily justify much lower estimate for every blockchain by taking more factors into consideration, but the 178 remains a handy guesstimate. Yes, other measures would look better for us. Imho our time would be better spent trying to deploy applications and attract users, rather than counting validators on the head of a pin.
At the same time, Nakamoto coefficient remains a handy criticism for chains who really fuck up their consensus and cannot even run 300-1000 validators, like Cosmos or ICP. Yet, the truth winds up much more complex, so if you take Nakamoto coefficient too seriously, and if you’re at all honest, then you’ll end up going nihilist and argue everybody has Nakamoto coefficient zero.
Anyways, if you see concrete community or security measures being taken by another ecosystem, then please do tell us. Jonas, Alistar, me, etc would think about if & how to integrate their ideas, like we’ve done many times before.
