7 ms·

If you're interested in the technical coordination part of it you may be interested in the client diversity ideal that Ethereum has. Changes to Ethereuem don't

If you're interested in the technical coordination part of it you may be interested in the client diversity ideal that Ethereum has. Changes to Ethereuem don't happen by implementing the change on a client, but by updating a spec and then the ~5 execution clients or the ~5 consensus clients all update their code to become compliant. It's a really high cost to pay, but in theory (assuming that the mix that is run is actually diverse) can protect against implementation bugs. In the browser world It'd be like running two different browser engines and only displaying DOMs that are consistent with both engines, or JS engine or final render etc.

kkielhofner 4y ago
It's a worthy ideal but it doesn't shape up in reality. 75% of current eth clients are one implementation[0]. Beacon is a little better but it looks like it could be heading that way (with Prysm already being well over half). I would have loved to contribute to client diversity but my experience with non-geth and non-prysm clients has been so bad I did what it looks like everyone else does - throw my hands up and jump on the bandwagon of what seems to actually work. [0] https://ethernodes.org/ https://ethernodes.org/
- zeroclip 4y ago
  Client diversity has improved in recent months, particularly consensus clients: https://clientdiversity.org/ https://clientdiversity.org/
  - kkielhofner 4y ago
    For beacon that is definitely better than the last time I looked. However geth is still 75% and relevant to the topic of the merge, roughly 88% of current clients aren't ready for it[0]!!! It will be interesting to see what these numbers look like when the merge actually happens... [0] - https://ethernodes.org/merge https://ethernodes.org/merge
    - michaelsbradley 4y ago
      The maintainers of geth recently published and then flagged (as bugged) the first release that was fully Merge-ready, so it's not really possible that geth operators could be ready at this time. I believe just today the next release was published, fixing the bug in the previous one: https://github.com/ethereum/go-ethereum/releases https://github.com/ethereum/go-ethereum/releases
      - kkielhofner 4y ago
        I covered that debacle in another comment: https://news.ycombinator.com/item?id=32581419 https://news.ycombinator.com/item?id=32581419 As noted there prysm, geth, etc have had their first merge-ready releases less than 48 hours ago (with one "oops" already) I (for one) will be waiting for what I'm sure will be more bugs to shake out in the next several days.
      - pcthrowaway 4y ago
        If geth isn't ready for the merge, how were (formerly) PoW testnets upgraded to PoS already?
        michaelsbradley 4y ago
        Because geth was ready for those test merges.
- X6S1x6Okd1st 4y ago
  Yup, one of the big problems is that MEV typically happens on geth right now so there's a huge financial imperative to use geth. MEV is another topic that is quite interesting from a technical & game theory point of view, but some parts of it are pretty horrifying from a moral point of view (e.g. front-running, forced liquidations, sandwhich attacks). Right now deciding what transactions go into a block (which includes manufacturing your own transactions that yield riskless profit (called MEV)) is tightly bound to running geth. Flash bots is working on ways to decouple that so there is increased specialization and allows for flexibility in which client you run.
- xur17 4y ago
  FWIW, ethernodes is not an accurate source of data on this. They are missing a LOT of nodes (mine has been running for over a year and isn't listed, and others I have spoken to say the same thing).
RjQoLCOSwiIKfpm 4y ago
Is it true that the Ethereum foundation itself doesn't develop a fully usable set of clients anymore for the post-Merge network? There seem to be only 3rd-party and seemingly commercial entities developing the "consensus" client which is needed to use the PoS network? As a user it would be rather "meh" to not have an official client to rely on :| See: "Ask HN: Does the Ethereum foundation really not develop a post-Merge client?" https://news.ycombinator.com/item?id=32586172 https://news.ycombinator.com/item?id=32586172 TL;DR: To use the new Proof of Stake network, you'll need 2 pieces of software in parallel ("execution client" and "consensus client"). It seems only one of those is developed by the Ethereum foundation - implementations of the other one are only developed by various seemingly commercial entities.

kkielhofner 4y ago

It's a worthy ideal but it doesn't shape up in reality. 75% of current eth clients are one implementation[0]. Beacon is a little better but it looks like it could be heading that way (with Prysm already being well over half). I would have loved to contribute to client diversity but my experience with non-geth and non-prysm clients has been so bad I did what it looks like everyone else does - throw my hands up and jump on the bandwagon of what seems to actually work. [0] https://ethernodes.org/ https://ethernodes.org/

zeroclip 4y ago
Client diversity has improved in recent months, particularly consensus clients: https://clientdiversity.org/ https://clientdiversity.org/
- kkielhofner 4y ago
  For beacon that is definitely better than the last time I looked. However geth is still 75% and relevant to the topic of the merge, roughly 88% of current clients aren't ready for it[0]!!! It will be interesting to see what these numbers look like when the merge actually happens... [0] - https://ethernodes.org/merge https://ethernodes.org/merge
  - michaelsbradley 4y ago
    The maintainers of geth recently published and then flagged (as bugged) the first release that was fully Merge-ready, so it's not really possible that geth operators could be ready at this time. I believe just today the next release was published, fixing the bug in the previous one: https://github.com/ethereum/go-ethereum/releases https://github.com/ethereum/go-ethereum/releases
    - kkielhofner 4y ago
      I covered that debacle in another comment: https://news.ycombinator.com/item?id=32581419 https://news.ycombinator.com/item?id=32581419 As noted there prysm, geth, etc have had their first merge-ready releases less than 48 hours ago (with one "oops" already) I (for one) will be waiting for what I'm sure will be more bugs to shake out in the next several days.
    - pcthrowaway 4y ago
      If geth isn't ready for the merge, how were (formerly) PoW testnets upgraded to PoS already?
      - michaelsbradley 4y ago
        Because geth was ready for those test merges.
X6S1x6Okd1st 4y ago
Yup, one of the big problems is that MEV typically happens on geth right now so there's a huge financial imperative to use geth. MEV is another topic that is quite interesting from a technical & game theory point of view, but some parts of it are pretty horrifying from a moral point of view (e.g. front-running, forced liquidations, sandwhich attacks). Right now deciding what transactions go into a block (which includes manufacturing your own transactions that yield riskless profit (called MEV)) is tightly bound to running geth. Flash bots is working on ways to decouple that so there is increased specialization and allows for flexibility in which client you run.
xur17 4y ago
FWIW, ethernodes is not an accurate source of data on this. They are missing a LOT of nodes (mine has been running for over a year and isn't listed, and others I have spoken to say the same thing).

RjQoLCOSwiIKfpm 4y ago

Is it true that the Ethereum foundation itself doesn't develop a fully usable set of clients anymore for the post-Merge network? There seem to be only 3rd-party and seemingly commercial entities developing the "consensus" client which is needed to use the PoS network? As a user it would be rather "meh" to not have an official client to rely on :| See: "Ask HN: Does the Ethereum foundation really not develop a post-Merge client?" https://news.ycombinator.com/item?id=32586172 https://news.ycombinator.com/item?id=32586172 TL;DR: To use the new Proof of Stake network, you'll need 2 pieces of software in parallel ("execution client" and "consensus client"). It seems only one of those is developed by the Ethereum foundation - implementations of the other one are only developed by various seemingly commercial entities.