vitalik.eth Profile picture
Oct 1 13 tweets 3 min read
The big really valuable and necessary thing that ERC-4337 provides for account abstraction is a *decentralized fee market* for user operations going into smart contract wallets.

eips.ethereum.org/EIPS/eip-4337
You should be able to send an op into a public mempool, and if it pays enough fees, reliably expect it to get included. This should NOT depend on ANY:

* Centralized actors
* Reputation systems for op senders
* ETH held in a separate EOA
* External services for account creation
This is the gold standard that the existing public mempool provides. And we absolutely need to provide the same guarantees for smart contract wallets, or else we risk worsening centralization and censorship.
Doing this securely requires a mempool node to be able to determine whether or not a given op is guaranteed to pay fees before rebroadcasting it.

This is quite hard to do, because it requires checking that the op does not depend on state that might change in the next block.
ERC-4337, and its predecessor EIP-2938, do this by enshrining a split between "validation" and "execution".

The validation step is bounded in how much gas it can consume, and cannot use "forbidden opcodes" that access state outside the account. If an op does this, it is rejected
All this ensures that if an op gets broadcasted through the mempool, then it is guaranteed that that op can get included on-chain and will pay fees - unless another op from that same account gets in first. Like Ethereum txs today.

All this prevents DoS attacks on the mempool.
The "paymaster" feature allows paymaster contracts, which let you separate "who pays for the op" and "who sends the op".
But if you use this feature, then you are relying on some external mechanism to pay for the op, and you should better be sure that it won't disappear on you!

(If it does, you would have to pay ETH and push the op through yourself)
The signature aggregation feature allows builders and batch submitters to also aggregate signatures (eg. BLS, later with SNARKs), greatly reducing the data that goes on chain. This is a critical cost-saving feature for rollups, where data costs dominate.

Now, problems that are NOT solved by ERC-4337!

* Letting existing EOAs send ops that get paid for by third parties
* Upgrading existing EOAs to contracts
* Integration with transaction inclusion lists (an anti-censorship feature in PBS)
These can't be done with an ERC. They require an actual consensus-changing EIP.

* EIP-3074 does the first
* EIP-5003 does the second
* No EIPs yet for the third, but see an introduction to the problem and a solution path here: notes.ethereum.org/@vbuterin/acco…
The good news is that we're narrowing down toward an actual path for account abstraction, something we've always wanted but have not managed to actually have!

And with the "carrot" of cost savings in rollups with signature aggregation, there's finally an incentive to adopt it.
Big thanks and appreciation to all the amazing people working on refining all of these proposals and actually getting them out there. Excited to keep moving this forward!

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with vitalik.eth

vitalik.eth Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @VitalikButerin

Sep 8
Second update from Balvi! 🧵

Lots more anti-covid projects funded, thanks to funds from @Shibtoken @CryptoRelief_!

Previous update here:

First, should you still care about covid?

I've seen the full spectrum of opinions from very smart people. I'd say it's not worth overturning our lives over it, but there are plenty of signs that this thing is a tier more dangerous than the cold/flu and we should not ignore it.
The biggest reason to worry about covid is not short-term death rates, but rather long covid. A few links:



mdpi.com/2076-0817/10/1…

science.org/content/articl…

axios.com/2022/08/25/lon…
Read 17 tweets
Sep 5
A thread of random polls about myself:
What D&D alignment most accurately describes me?
What D&D alignment most accurately describes me?
Read 13 tweets
Sep 4
Neural networks as philosophy:

Late 20th century philosophical thought was heavily influenced by intuitions from study of computers and algorithms: *we too* are algorithms, what could those insights say about us?

I wonder if the 21st century will do the same with neural nets.
When people say things like "human beings are not algorithms", sometimes they're not literally asserting a spirit world not describable with math and physics, rather they're saying: we're different from *computer programs written by humans*. We're more similar to neural nets.
Computer programs are highly structured, they are legible, and they are either perfect or have bugs that break things horribly until fixed.

Neural nets are highly illegible, they make mistakes, but somehow even mistaken answers are... kinda reasonable? Image
Read 4 tweets
Aug 4
I continue to find probability distributions with infinite means fascinating. Particularly how their averages *don't converge* as you sample more values.

You can even replicate the behavior deterministically with "tan at integers" as a pseudo-RNG.
And if you use tan^2 you get into total crazy land.
What's even more interesting is that tan(int) is *not* a random number generator. There's fascinating patterns in the data.

One way to look at it is: if tan(x) is very high, x is very close to a 90' or 270' angle. So if tan(x) and tan(y) are high, tan(x + 2y) is also high.
Read 4 tweets
Jul 7
I just remembered a conversation with @AnthonyLeeZhang from pre-covid times on the role of ideology in an intellectual's thinking.

The dichotomy of "idea-driven ideas" vs "data-driven ideas". Framed this way, ideology looks bad, but actually it's not so simple...
... because there are other options. A particularly important option is "data-driven choice of idea-driven ideas". Ideology generates hypotheses and provides motivation, but then you check them against reality, and be prepared to dismiss or de-emphasize some as a result.
Motivation truly is important. You can keep going longer and focus harder on the mission if you can internally narrativize your intellectual life as a glorious search for freedom, enfranchisement of the weak, human progress, etc. Such energies can't be thrown away completely.
Read 5 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

:(