Sepolia beacon chain

[djrtwo]

EDITED

from @parithosh:

As discussed on ACD, Sepolia will have a minimal validator set. I'd say 100 validator keys per entity taking part (Assuming that every client team wants to take part) => Validator set of 1212 (9 client teams + EF + Attestant/Vouch + 12 for EF security + Alex).

Having a validator set of 100 per entity means that we can assign low number of keys (like 12) to entities such as EF security without them representing a significant portion of the network.

If there are entities outside of client teams that would like to take part, please comment below.

[q9f]

Did you discuss the option to have a small, permissioned validator set, i.e., something along the lines MIN_GENESIS_ACTIVE_VALIDATOR_COUNT = 64 and a deposit contract that rejects new deposits afterward?

We should have one testnet that allows developers to expect a fairly stable application-test environment for dapp and contract developers (Sepolia?) whereas we have another one that can be used for consensus and execution client teams to test protocol mechanics (Goerli/Prater?). A small validator set would reduce the friction to maintain keys and validators long-term (IMO).

[parithosh]

We considered a small validator set for Sepolia but didn't want to miss the chance to have a new beaconchain with a large genesis set. The main reason is that in order to expand Prater to 2xmainnet we'd need to perform a non-insigificant number of deposits with an increasingly reducing supply of goerli ether. By making the Sepolia beaconchain much larger, we can stop trying to increase the validator set on Prater and reserve that goerli ether for other purposes.

If we're fine continuing to increase the prater validator set size, then I agree that we can have a small network for sepolia that's a lot easier to run.

[parithosh]

Pro/Cons of various beaconchain sizes:

Large beaconchain (2x mainnet):
Pro:

• Easiest to bootstrap right at genesis
• We can use a token contract to save sepolia eth from being wasted
• Able to test a far higher load and be ahead of mainnet more easily

Cons:

• Client teams need to run a lot of validators on both prater and sepolia-beaconchain

Small beaconchain
Pro:

• Simpler to setup, easier on client teams
• Just one large network (prater) with all client teams running significant validators

Cons:

• It would mean we'd need to continue making deposits on prater to catch up to mainnet => Spend a lot of goerli eth

[parithosh]

As discussed on ACD, Sepolia will have a minimal validator set. I'd say 100 validator keys per entity taking part (Assuming that every client team wants to take part) => Validator set of 1212 (9 client teams + EF + Attestant/Vouch + 12 for EF security + Alex).

Having a validator set of 100 per entity means that we can assign low number of keys (like 12) to entities such as EF security without them representing a significant portion of the network.

If there are entities outside of client teams that would like to take part, please comment below.

[q9f]

If there are entities outside of client teams that would like to take part, please comment below.

I'd be happy to participate on behalf of the Goerli community.

[timbeiko]

On #528 we agreed that we should launch the sepolia beacon chain as soon as (reasonably) possible, and to confirm this with CL teams on the next CL call.

[lightclient]

Just wanted to leave a comment to say that I'm setting up a some infra to run a lot of validators on testnets (starting with prater) and would also be happy to also run on sepolia.

[nalepae]

If there are entities outside of client teams that would like to take part, please comment below.

We would be happy to participate too.

[infosecual]

I just read the CL Call#88 notes and noticed that the sepolia beaconchain "will be permissioned, with ~2k validators." (link)

Aside from the 12 validators that the EF CL Security Research team is slated to get we were planning to run other validators to test various things like slashing. We have been mining sepolia ETH for the last few days to seed this other set of validators. If the validator set will be permissioned I would like to request that we get more than 12 validators (maybe ~50) or that there is some method provided for us to deposit our mined SEP in a permissioned manner.

The plan is to have 12 honest, longer term validators that are running various sanitizers (MSAN, ASAN, UBSAN, TSAN, etc.) and to have others that will either be slashed on purpose or otherwise not reliable in some fashion.

[q9f]

The plan is to have 12 honest, longer term validators that are running various sanitizers (MSAN, ASAN, UBSAN, TSAN, etc.) and to have others that will either be slashed on purpose or otherwise not reliable in some fashion.

Please use Goerli/Prater for this.

eth-clients/goerli#98

[parithosh]

I think the idea is that they'd do this for each testnet, so they'd do this for goerli as well :)