Hacker News Reader

Comment by kkielhofner | Hacker News Reader

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/

View on HN

Client diversity has improved in recent months, particularly consensus clients:

https://clientdiversity.org/

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

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

I covered that debacle in another comment:

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.

If geth isn't ready for the merge, how were (formerly) PoW testnets upgraded to PoS already?

Because geth was ready for those test merges.

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.

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).

Reply on news.ycombinator.com