Deprecate this repo, move to per-network repos

[arnetheduck]

Given how sepolia and görli already have their own repos in the same format as clients use for loading genesis etc, I'd suggest the following changes given that the rest of the information in this repo in particular is mostly obsolete:

• create a new mainnet project (https://github.com/eth-clients/mainnet probably)
• move the current state of shared/mainnet there
• mark this repository archived

[q9f]

I agree, it makes sense to have long-standing, public networks in their own, contained repositories. sepolia is currently maintained in two places: eth-clients/merge-testnets#22

for mainnet, we should consider hosting it along the spec though, or maybe under the ethereum github org.

[arnetheduck]

for mainnet, we should consider hosting it along the spec though, or maybe under the ethereum github org.

I'm about to suggest the opposite direction in the spec repo: remove the instance-specific variables in the spec examples (ie fork epochs) and replace them with dummy values (ie FAR_FUTURE_EPOCH) - the spec is generic, so I feel it's a bit odd to special-case the mainnet "instance" - also, there is ephemeral information in the configuration such as boot nodes as well as large cached binary blobs (genesis) which I think generally don't belong in a spec repo that describes a protocol.

I believe the mainnet repo is / should be a shared responsibility between the the EF and client teams (which is why eth-clients was set up) - the spec is already under EF control meaning there is no ambiguity in terms of what we're all implementing whereas a chain definition contains the information we want clients to agree on before launching into a hard fork and that largely are maintained by client teams (ie fork epochs, boot nodes etc) - I see it as a net positive that this repo is owned by a wider range of actors.

[parithosh]

We could move to a folder structure similar to what we have here: https://github.com/ethpandaops/dencun-devnets/tree/master/network-configs/devnet-12

That's the folder we use for every client out there in testnets, would definitely be happy if some of the redundant files get deleted in the process.

[arnetheduck]

We could move to a folder structure similar to what we have her

Well, I'm thinking mostly about what to call the top-level folder, ie holesky calls this custom_config_data in sepolia it's bepolia and your link has .. well, the testnet name but that's because it's a multi-testnet repo (vs a single-network repo).

I think if we have something like:

README.md
config/
  genesis.ssz
  etc

you can reuse that for testnets as well, ie you'd have:

testnet-1/
  config/
  tranches/
  etc...
testnet-2/
  config/
  tranches/
  etc...

tranches in particular are of no interest to clients while the files in config are (we load genesis from there, as well as bootnodes and chain parameters - it's nice to have these separated from ephemeral-testnet stuff... wdyt?

[protolambda]

We used to have https://github.com/eth-clients/eth2-mainnet maybe it's time to revive that repo, rather than starting a new one?

[arnetheduck]

sadly, eth2 has fallen out of favor - fine with me but ..

[parithosh]

We could move to a folder structure similar to what we have her

Well, I'm thinking mostly about what to call the top-level folder, ie holesky calls this custom_config_data in sepolia it's bepolia and your link has .. well, the testnet name but that's because it's a multi-testnet repo (vs a single-network repo).

I think if we have something like:

README.md
config/
  genesis.ssz
  etc

you can reuse that for testnets as well, ie you'd have:

testnet-1/
  config/
  tranches/
  etc...
testnet-2/
  config/
  tranches/
  etc...

tranches in particular are of no interest to clients while the files in config are (we load genesis from there, as well as bootnodes and chain parameters - it's nice to have these separated from ephemeral-testnet stuff... wdyt?

I think we can leave traches out of it and just include the validator indexes of known genesis entities in the README. I like the idea of having the format as, this would also work with all of our tooling:

README.md
config/
  genesis.ssz
  etc

re: eth2-mainnet, Maybe we can just rename the repo to mainnet and continue using it? Although afaik mainnet configs are derived from the consensus-specs repo anyway, so im not sure if it provides any extra value.

[mratsim]

network-mainnet, network-holesky, network-..., would help with discoverability and having all next to each other when sorting.

Alternatively, we rename the repo eth-networks and put all network config (execution and consensus) in this repo.
Sepolia and Goerli are being deprecated soon anyway.

[arnetheduck]

Maybe we can just rename the repo to mainnet and continue using it?

I don't quite see the usefulness - we'll get some stale git commit history, but apart from that, not much - a clean repo will have a clean history