> I think you know my argument is that it’s all tied together

Well it’s not. “It’s all tied together” is the sort of argument that’s generally made as an attempt to defend unconnected idiosyncratic beliefs of a political tribe, not neutrally trying to figure out the best parameter along each dimension. I personally think it’s very important to guard against that kind of argumentation.

The object-level issues right now are:

• How harmful (or beneficial?) to decentralization is having a governance model that makes it easier to change the difficulty of reading the blockchain?
• How harmful to decentralization is sharding?
• How beneficial or harmful to decentralization is proof of stake, as opposed to proof of work?

One is dx/dd, the other is dy/dd, the last is dz/dd. Sure, the d (“decentralization”) is what we’re going after and is the strand that ties them together, but those are three separate slopes, with very different and unconnected arguments about each one.

> which brings about the Dapp argument again because you know they’ll get priced out.

“If you don’t raise the blocksize, stuff will get priced out” is an argument that applies to money transactions as much as it does to dapp activity. Now you can say, “layer 2 will save the day!”, and we can also say “layer 2 will save the day” until sharding is ready; we kind of are with the recent emphasis on state channels and Plasma.

> Ethereum nodes don’t look and say “that’s over X, nope not valid”, there’s a process (of which I, again, already explained earlier than this quote) that effectively let’s those in control set it to whatever they like with enough time to bring the average up.

The gas limit is adjustable via on-chain governance via miners, and a 51%+ coalition has the ability to adjust it to whatever value they want. Yes, I get it. A 51% coalition also has the power to revert the blockchain, censor arbitrary accounts, and steal money from arbitrary channels or other layer-2 games. Incidentally, it’s PoS that does a better job of giving users more power by giving them a better array of responses to such situations via the minority soft-fork mechanism, and by having far lower barriers to entry to participation to make 51% attacks less likely…. but oh well.

> My Bitcoin node isn’t just “checking”. It’s refusing to send to it’s peers invalid information, and block peers that continuously try to send invalid information. Although I made this pretty clear in the section on Bitcoin’s network, they all validate and then propagate. Their collective power comes from them refusing the propagation of invalid blocks, and it gets stronger as the set of them grow.

This seems like a dangerous thing to rely on. Fundamentally you seem to be saying that non-mining full nodes actually contribute to network security in an active sense, because they insert themselves between other nodes and refuse to propagate blocks that they deem invalid. However, if this power were actually significant, it could easily be abused. A malicious actor (say, BitPay) could cheaply acquire a 90% share of full nodes by number in the network, and then refuse to propagate blocks that contained transactions from some party they wanted to inconvenience (say, Coinbase). If this power mattered, it would become a nasty censorship vector. Fortunately, this functionality does not really matter, because you only need a few connections to broadcast blocks successfully and miners have relay networks anyway, etc etc, but then if that’s true then it’s also true that full nodes relaying all blocks that pass basic PoW checks would not affect decentralization much either. A full node’s value to the network is only in its implicit contribution to the coordination difficulty of pushing an undesired hard fork onto the network, not in weird p2p stuff.

FWIW, I’ve also never heard Bitcoin proponents make this argument about the p2p network before. https://medium.com/@lopp/securing-your-financial-sovereignty-3af6fe834603 for example focuses on privacy, full validation for personal benefit, and being able to serve light nodes. https://medium.com/colu/full-nodes-and-fake-news-a-bitcoin-primer-for-bitcoiners-47120b1a97bf also focuses on something fairly similar to my coordination argument. So I’m not sure even they would agree that full nodes not broadcasting blocks through the p2p network is key to security.

Third, in the case of sharding, this exact same security model exists. There’s a subnet for every shard, and nodes that are currently in that shard subnet refuse to propagate blocks in that shard that are not valid.

> It seems like you’re using “PoS Validator” in a way that to me means “Block Creator”, and then mirroring that over to Bitcoin’s network and calling “nodes that happen to mine” “PoW Validators”. All nodes are equal in Bitcoin, and while they may also be equal in Ethereum with Proof of Stake, they aren’t with Sharding.

This needs more justification. I still fail to see the sense in which you’re claiming that nodes in sharding suddenly become unequal.

> What I’m actually doing is making a case for why Bitcoin is better on both fronts, and how they need to go hand in hand. More importantly, I think “everyone as a full node” ideal is directly hindered by the miner-adjustable blocksize, so the debates must join together.

See above point about dx/dd and dy/dd.

> Light nodes can check, but they have no say. They are literally a tier below the network of validators (my definition, not yours). When I say “these are the nodes you’ll be running”, it’s true because they won’t be staking 32 ETH.

The Ethereum network (as in, the p2p network) also makes no distinction whatsoever between staking and non-staking nodes. Staking is just a function that some nodes happen to do, much like mining is just a function that some nodes happen to do. The p2p layer makes absolutely no distinction here; all it sees are nodes that validate the beacon chain and hop around between shard subnets, some of which do so because they’re incentivized to by the staking process and others do so for whatever other personal reasons.

> If those two things were true, barring the complications over keeping them true you’d still have the issue where fees will outbid each other and Dapps dependent on low fee transactions (most of them) become useless.

Yes, there’s a tradeoff between predictability of reading cost and predictability of writing cost. We’ve been through this. Sharding improves the tradeoff frontier by maybe a factor of 1000; other technical improvements can probably achieve another 10x and layer 2 systems can achieve further gains of hopefully 3 orders of magnitude, and that’s how we get to the scalability levels that we need for decentralized applications to work well for large-scale use.

Source and More information: > I think you know my argument is that it’s all tied together

Author: Vitalik Buterin