Seeing a lot of talk about Terra lately, which runs on Cosmos.
Given that Cosmos limits the validator set to 100, do people really expect a system limited to such a tiny number of validators to successfully secure hundreds of billions or trillions in the long run?
Conversation
This is a really good question, and something I spend a lot of time thinking about.
Tendermint is perhaps the simplest possible form of PoS - theres nothing special about it that constrains it to a smaller validator set vs other PoS systems including eth 2.0
4
3
31
Cosmos chains take a DPoS flavor on top of Tendermint, and the philosophy is to prioritize decentralization among a small set of delegates to achieve max efficiency in on-chain governance and responsiveness to events
1
9
All PoS systems have theoretical limits on how far the validator set can scale per shard before performance starts to degrade significantly - so the only options are to scale vertically (increase hardware specs) or horizontally (segregating state machines in shards)
1
9
In some sense, the hub and spoke design of Cosmos / IBC is intended to horizontally scale in "branded shards" - each cosmos state machine has a sovereign set of rules, but they all speak a common language to sync important state via IBC
1
11
So in some sense, horizontal scaling through branded shards in cosmos is already feature complete - the new challenges will be, how can you build apps that are agnostic to the shards that underly them? I.e. now that you've scaled the machine, how can you scale apps?
1
2
15
In the beginning, developer ergonomics is going to be a bit janky - apps may need to "be aware" of what state lives in which zones - but it's not a very difficult problem to eventually abstract this away
1
7
Show replies
I think adoption of asynchronous programming models on top of IBC (think async/await in JS/Rust) is going to unlock a lot of usecases w/o resorting to shared security
4