Barnabas Busa

@josephdelong @colossuspay 👀 need a hand?

Who up deploying their testnet?

@peter_szilagyi 🤣🤣🤣

Yeah, now I understand why I switched over to USB-C :P Alas, I don't have enclosures for that yet.

Sigh, I wanted to prep 2 alpha Dark Bio Arks to give away on Tuesday. Managed to burn *both* of them due the magnetic connector. It snapped on slightly shifted and still made contact, but 1 pin shifted. USB data controller instantly gone.

@keoneHD @raulvk Bandwidth scaling is fine, agreed. But O(N) fan-out isn’t a bandwidth problem, it’s a connectivity problem. 10k validators each maintaining 9999 peers is a full mesh. Ethereum gossip works because each node talks to ~75 peers. Tested beyond 200 validators?

This isn't true!! This is the beauty of RaptorCast. To recap, RaptorCast follows a two-level broadcast tree (with assignments of chunks to the first level weighted by stake) with erasure coding (with 2.5x redundancy factor). Let there be V validators, and let B be the size of the block in bytes. In RaptorCast's two-hop broadcast tree: - in the first hop, the leader sends roughly 2.5*B bytes of data, spread across all other validators, i.e. each validator on average gets 2.5*B/V - in the second hop, each validator sends the chunks that it receives to every other validator. So the total amount of data it needs to send out (on average) is 2.5*B/V * V = 2.5*B. For example, with 10k nodes with (for simplicity) equal stake weight, each validator is the first-hop validator for 2.5/10000 = 1/4000 of a block worth of chunks. Then in the second hop, each validator sends 1/4000 of a block worth of chunks to 10k other validators, i.e. it sends 2.5 blocks worth of data. The more immediate bottleneck for 10k validators is cryptography (certain operations in MonadBFT (our consensus) have costs that scale with number of votes) but there's active work being done there. But the point of this post is to say that, on the p2p layer, this problem is largely addressed by RaptorCast. We can have performance with decentralization. Longer blog post here: category.xyz/blogs/raptorca…

One thing people don’t realise is that scaling a 10k node decentralized p2p network with ~1MM active validators is a diametrically different problem to scaling a federated cluster of ~172 active validators, hard-capped by design to 200. Ethereum is designed to be universally accessible infra: any person can plug and play a commodity machine to a commodity network and become a validator with 32 ETH. A NUC, a Raspberry Pi, a VM on cloud, it doesn’t care. Monad’s validators must run on bare metal, no VM, no cloud. Eth operates on 100/50 Mbps bandwidth; Monad requires 300/300 Mbps. Achieving fast finality across a few hundred validators is a well-understood problem. The hard problem is doing it across hundreds of thousands of globally distributed validators on heterogeneous infra. To put it in context, just the Barcelona to New Zealand RTT is ~300ms, that’s 3/4 of the Monad block time spent only on beaming photons across before any compute. Scaling such a network is a fundamentally different challenge. Yes, there’s a lot to optimize (much of it identified already), but at some point we’re up against physics and the speed of light over fiber. I could stand up a cluster on AWS tomorrow with 50 nodes and achieve higher throughput than Monad. But it makes zero sense to compare. Different architectures, different goals, different tradeoffs.

@Uniswap why do I have to sign in to read some docs?

Try it out ↓ developers.uniswap.org

Today we're opening up the Uniswap Developer Platform in beta Any builder can generate Uniswap API keys to add swap and LP functionality directly to an app in minutes Just in time for anyone hacking at ETHDenver 🫣

@potuz_eth @nero_eth From tooling perspective it’s also a very invasive change. Lots of our assumptions is based on a static slot time. I can only imagine how many app devs are also relying on hardcoded 12s somewhere in their code. This is going to be a significant and invasive change for public too

@nero_eth I don't know, but I am willing to bet that *no* client supports changing the slot time on the flight. Prysm had a prototype and took an invasive step to write. Yes **all** clients can start with shorter slot times, we all test Kurtosis on shorter slots. That's useless.

FWIW, I'd love to see CL devs chiming in on whether this is a headliner or not. The approach of shipping the *shortest slot we can achieve without giving up (geographical) decentralization* is very appealing to me. I just was under the impression that any slot time change (including making it longer) would be a headliner-type PR both in implementation and scope. If this assessment is incorrect I'm happy to support this data-driven EIP alongside another headliner. A commitment to this EIP is also very simple as long as we don't commit to timelines.

@nero_eth @jih2nn @MorphNetrunner @nero_eth what are the prysm/geth branches ?

6-second slots are fully spec’d, prototyped in Geth and Prysm, and ready to ship next year. Easily one of the most impactful upgrades we can deliver in the Heka/Bogotá fork. h/t @jih2nn and @MorphNetrunner for getting a first local devnet running.

I’m a European citizen. I’m not a proud one today. When I was growing up I was told that the strongest economy the smartest people and the most innovative people were European. While that might have been true for the 20th century, it sure isn’t true for the 21st century. Every step of the way Europe gets left behind other nations, and not just a little bit. Change has to happen, deregulation has to happen, and innovation must be coming back to Europe if we want to stay relevant. I don’t think being leaders of anything is going to be possible, I’m just asking to be able to stay relevant. Energy is the most important factor when it comes to economic growth. Without energy growth will halt. As we are progressing towards a new age of energy consumer mammoth, the most important question that every nation should be asking themselves is how can we: - improve our energy output - improve efficiency how we move energy around - improve our energy storage - reduce the amount of time it takes to do all the steps above There is no silver bullet for high output, low cost energy. I believe the biggest blockers currently in the European energy sector is the insane amount of regulation that corporations have to jump through just to be able to become energy providers or consumers. “Feasibility studies” should take hours instead of months. Communication across the globe takes milliseconds, why is the feasibility of whether a new 5-50MW power plant can come online is not obvious right away. If you need to add new lines to make it happen, then let the investor company know and who knows you might actually get helped out. But it shouldn’t take months just to find out that something is or isn’t possible. Countries should commercialize the energy grids and let corporations operate the grid, while let governments oversee fair trade. Allow small businesses to install battery solution, introduce tax breaks if need to, to further stabilize the grid. Push for reforms where solar parks are encouraged to store their power for peaks and not feed back to the grid when demand is low. Currently it takes 6 months to 1 year - 400 pages of regulatory BS - to install a battery backup to an existing 0.5MW solar park in Hungary. This is behind the meter, no grid effect (other than benefit). Plan+execution of given battery takes about a week. While I write this post, in China another 0.5/1MWh capacity has just been installed (estimated to about 200MW/400MWh being installed every single day) This is the reality we are up against. You want cheap electricity and stable grid but the cost of doing business in terms of bureaucracy overhead kills every new startup idea and innovation. I’m calling out every nation in europe to: - Increase AC capacity like it’s the industrial revolution. Install new capacity like there is no tomorrow. - Increase transformer capacity at all your substations. If it’s maxed out, build new ones. - Reduce the overhead in regulation for every small project. Cut down response times from 6mo to 6 days at most. If the architecture plan and execution can be done in under a week, paper pushing shouldn’t take that long. - Don’t make people go to 15 different governmental regulation body for paperwork. Energy sector should be handled by a corporation, not by government. It’s in this corporation’s best interest to improve the quality of services it can provide. Adding a new service is in the best interest of the grid operator, as it will help balance the load or help generate revenue. - Adding new producer capacity should be easy. - Buying from existing capacity should also be just as easy. Innovation has left Europe, we are playing catch up to giants. Let’s at least try to stay relevant. @euaccofficial

@parithosh_j @christine_dkim I’m interpreting the spec to the best of my knowledge.

@christine_dkim @BarnabasBusa

meme of the week from yesterday's ACD After Hours newsletter cc: @BarnabasBusa 😁

@gakonst how would you ship features that are changing the consensus mechanism without having multiple forks on the live network?

Why should Ethereum upgrades happen in big bundled 'hardforks' where node operators are forced to upgrade their nodes by a date? Why shouldn't nodes gate new features & only activate them after X% of stake signals activaton? Would allow us to move much faster.

EthPandaOps pivots to new challenges. We think GitChain is the way forward. To learn more check out our latest blog post below. ethpandaops.io/posts/april-fo…

@777ethgrind @parithosh_j @SwaroopH @anomit @xadahiya houdini

Lore: Hoodi is a metro station in Bengaluru, India. The canonical pronunciation is ಹೂಡಿ (hoo-dee i guess?) Mascot is a peacock in a hoodie :D

@lightclients @marek_ @coinbase do they need a hand?

@marek_ @coinbase why are they not? is it just lazy devops?

It's disheartening that @Coinbase has chosen not to support the Celo L2 upgrade. This feels like a wrench in Ethereum's layer-2-centric scaling roadmap. Why would other EVM-compatible L1s follow suit now?

@thebensams @peter_szilagyi @parithosh_j @0xemon @onbjerg @_pk910_ already at Aquilo

@peter_szilagyi @parithosh_j @0xemon @BarnabasBusa @onbjerg @_pk910_ just wait until you land on gelba, then the factory becomes a 24/7 obligation

Rust is the future of programming. Me: Fun, how do I write concurrent stuff? Rust: Use Tokio! Me: Cool, what is Tokio? Rust: It's a lib that implements Go concurrency =)

@0xemon @peter_szilagyi @parithosh_j @onbjerg @_pk910_ Can cursor grow the factory?

To the best team one could ask for <3

@remy_roy @peter_szilagyi yes

@peter_szilagyi Main bus or spaghetti?

The alternative is Factorio Space Age :/... decisions...

@parithosh_j All but 7742 (which shouldn't be user facing anyways).

If you want something to hack on at DevCon, try out the Mekong testnet! All the pectra EIPs are live: so you can brainstorm your EIP-7702 wallet user flows, deposit a validator to use MaxEB and exit a staking pool validator with a smart contract!

@peter_szilagyi you can already do this with kurtosis with k8s backend.

Can anyone point me to best practises wrt cross-node/cross pod P2P networking in k8s? If I were to deploy something based on libp2p, how painful would it be to have a cluster that can talk between each other directly instead of some hierarchical topology?

@prestonvanloon Also cutting edge hw. Great example from @gballet compiling geth in risc-v then pushing it to [dockerhub](hub.docker.com/layers/gballet…) Can prysm run on risc-v?

Ethereum L1 should support middle market consumer hardware for 5 years. In other words, Ethereum remains accessible for reasonably aged hardware. This allows home users/stakers to use older hardware and upgrade every few years. Supporting the last two MBP models (high end market hardware) or recent enterprise/server grade hardware isn't appropriate for a sufficiently decentralized network.

@lodestar_eth `disableKeystoresThreadPool` (note: pls don't use it on your mainnet validators)