The Internet Computer in a world of modular blockchains

How will it adapt to this reality?

One example apart from Eth would be: https://celestia.org/ but if you check the whole thread, its far from the only one

I don’t quite get the whole “modular blockchain” thing.

Also what makes Ethereum not monolithic? Because it has rollups? The IC will soon integrate with BTC and later this year with ETH. Does that make it still monolithic? In fact, the IC is really a collection of many subnet blockchains that are governed by a single “master” NNS blockchain. Does that make it modular?

If the issue is more about validity / fraud proofs (which I think are variations of “zero knowledge proofs”), then I admit I’m not an expert at all.

Can someone explain what’s the argument for every blockchain needing zero knowledge proofs or risk becoming irrelevant? My understanding is that zk proofs provide additional security guarantees on top of BFT consensus. But they are more of a cherry on top rather than a necessity. In Ethereum’s case, I suppose they are more necessary due to the lack of consensus sharing between the Ethereum blockchain and the various rollups, so they use zk proofs to “bridge” that gap. But why does a L1 like ICP need zero knowledge proofs, if we don’t even use rollups? Chain key technology seems to work fine as is…

4 Likes

ethereum is monolithic currently, but the plan is to move away from that.

Yes. because execution still happens on layer 1(s).
You can find some more technical explanations (am not an expert) but the basic idea is optimizing layer 1 for data availability & settlement and doing execution on layer 2s.

Well you dont necessarily need zk proofs. fraud proofs (optimistic rollups) may work as well.
but zk rollups when reaching their full potential could give insane amounts of tps (in the hundreds of thousands).

The idea is you get almost the same insanely high security guarantees as layer 1 while also increasing scalability by an order of magnitude.

Theres a lot of nuances of course, but I still think this question is worth asking as many projects are starting to look into that direction (not just ethereum).

On the IC all the ZK-stuff could be implemented on top of the application level right now. Since no-one is working in that direction (publicly, I mean), it seems like there is just no real interest within the community to do that. Which for me personally is a signal, that speed and lower costs (scalability, basically) are the real reasons for people to like L2 on Eth.

Since there is no problem with speed or costs on the IC (it can already scale infinitely, there is just no better option) - there is no L2.

If somewhen in the future there will be any good reason to have ZKs on lower levels - I’m sure, this could be done by adding another subnet type.

On the IC one could think of another type of L2 solutions, like ultra-cheap (but less resilient) storage or computational networks. Imagine that you have a network of nodes, against which you could run a MMO game server with thousands of users. Such a network could rely on ICs security guarantees, but provide real-time processing capabilities to its users.

4 Likes

I definitely agree speed is currently valued higher than security, and Im definitely not as alarmist about this as the tweeter is. Think it will take some time to play out.

However I’m not sure it would be accurate to call IC scaling “infinite” as it currently stands.
Since subnets don’t share security it really doesnt differ all that much from cosmos IBC aside from everything being more tightly integrated.

Even query calls have to go through consensus as it stands.

Is it fair to call cosmos scaling infinite? I dont think so.

Yeah could be something, but tbh Im more interested in the IC becoming a layer 2 (yeah yeah I know, not gonna happen) because I do think similar systems to the IC will sooner or later start popping up in other ecosystems, in the form of layer 2s.

Also the question then becomes, why even run it on IC if you get much higher security guarantees & other benefits with a layer 1 fully optimized for this (like celestia)?

1 Like

Here is an interesting thread that discusses how the IC could benefit from zk proofs.

My takeaway is that zk proofs enhance security and can enable novel single node trustless computations, but these use cases seem quite niche at the moment.

They play a more prominent role in the Ethereum ecosystem, because they seem to be used as a substitute for subnets + chain key in order to scale.

Of course, zk proofs are just one type of validity proofs. According to this post, the IC’s native threshold signatures are another type.

In short, I feel like other blockchains are playing catch-up but who knows…

But you can grow individual subnets infinitely, so it is infinitely scalable.

Also isn’t Cosmos IBC just a mechanism to wrap and transfer tokens between blockchains? Can your code in one blockchain call methods from code in another blockchain? I don’t think so.

1 Like

This would be like calling Bitcoin scalable because other blockchains exist and are connected to it.

Cosmos IBC isnt limited to just token transfers, thats just one of the first use cases.

The IBC application layer can be used to build a wide range of cross-chain applications, including but not limited to token transfers, interchain accounts (delegate calls between two chains), non-fungible token transfers and oracle data feeds.

Other blockchains are playing catch up in some areas, yes. UX & DX in particular are fantastic. But there’s other exciting developments happening outside the IC that can’t just be ignored.

Zk-proofs are not mandatory. Im purely speaking of layer 2s here (either zk proofs or fraud proofs may be used).

Even if chain key + subnets were a substitute (lets ignore the lack of shared security) I would still say its only half true. BigMap? Cross subnet queries without the latency?

Also see dank sharding (https://twitter.com/dankrad/status/1475995526478843905)

What blockchains exist that are trustlessly and highly securely connected to Bitcoin?

From what I understand IC does have global verification/security A Brief Tour of Dfinity - YouTube

Anyway it’s pointless to compare traditional L1s to IC because they fundamentally different platforms. They both use blockchain technology and smart contracts but for fundamentally different purposes. Ethereum/Solana/Cardano/Avalanche/BSC are decentralized databases. ICP is a decentralized cloud computing platform. It’s like comparing Oracle to AWS, apples and oranges. Ethereum ect… uses blockchain tech to utilize a network of nodes for for security first and compute capacity second. IC uses blockchain tech for compute capacity first and security second.

IC is not trying to be an absolutely secure immutable database. It’s more on the execution layer side of things. I’m more concerned about IC’s ability to compete with Rollups, who only have to focus on execution whereas IC has to worry about consensus/security as well. How capable will rollups be long term? Can their transaction speed/finality be faster than ICs 1-2 second update speed because rollups dont have to go through consensus and basically instantly finalize. Can rollups store data as cheap or more cheaply than IC? Will rollups have composabilty with eachother and interoperate as efficiently as IC subnets? Can ETH rollups run any VMs other than EVM? Will rollups be able to copy ICs distributed key generation and threshold signatures and box out IC of it’s main innovation?

How does IBCs protocal trustlessly and securely communicate? From what I think I heard/understand it basically used multisig wallets to interoperate it’s blockchains. That’s not very secure if that’s the case, but I’m not very familiar.

2 Likes

Im sorry theres too many questions here that have been answered already. But theres one legitimate thing you bring up:

Yes, chain key is great. The communication being secure is certainly one aspect. However…
You are still fundamentally trusting two completely separate blockchains (which subnets are) with a different set of validators. They do not share security and chain key doesnt make a difference to that fact.

1 Like

Yes they do share security A Brief Tour of Dfinity - YouTube

That is a talk from 2016 and its not accurate anymore. Please, you can find the answer to this literally everywhere.
Im actually trying to be helpful here so that we can secure the ICs future.

2 Likes

So you’re telling me Dominic Williams went from saying separate chains without shared security is “rubbish” and “you NEED to have global security” to implementing the exact opposite of that? I find that hard to believe.

You dont have to believe me. The image is from A Technical Overview of the Internet Computer | DFINITY | The Internet Computer Review but you can find the same info in other places if you read the documentation. As well as here on the forum from DFINITY employees.
For whatever reason they failed delivering it. Its not an easy problem exactly.
Continue in DMs please if you wish, dont want to derail the topic any further.

Keep in mind Dom also said canisters wouldn’t have to comply with state law cause they “lived” in cyberspace and we all know what happened with Mario64, so take old statements with a grain of salt.

Dom today says subnets have globally manged security https://twitter.com/dominic_w/status/1501557345293328385

“globally managed” does not mean shared security. hes referring to “deterministic decentralization” aka nodes being assigned by the NNS according to geography and other factors. Completely different thing.

Think about it. Why would different subnets have different node counts if security were global.
Separate topic please.

Why is this a separate topic? We’re talking about whether the IC is modular or monolithic.

How capable will rollups be long term? Can their transaction speed/finality be faster than ICs 1-2 second update speed because rollups dont have to go through consensus and basically instantly finalize. Can rollups store data as cheap or more cheaply than IC? Will rollups have composabilty with eachother and interoperate as efficiently as IC subnets?

If rollups don’t need consensus, then why bother with rollups at all? Why can’t everyone run their own computation on their own computer, and publish their own zk proof using some well-regarded open source library?

I’m not familiar with rollups tbh… this has been on my reading list for a while.

(Also, rollups shouldn’t be able to store data securely; as you said, they are an execution layer, not a data layer.)

1 Like