Ethereum Core Devs Meeting 96 Agenda
Souptacular opened this issue · 29 comments
Ethereum Core Devs Meeting 96 Agenda
- Meeting Date/Time: Friday 18 September 2020, 14:00 UTC
- Meeting Duration 1.5 hours
- YouTube Live Stream
Agenda
- EIP & Upgrades Updates
a. YOLO / YOLOv2 & Berlin state tests update
b. Breakout room feedback - EIP Discussion
a. EIP-1057 Next Steps.
b. EIP- BLS12-381 curve operations & BLS12-377 curve operations
c. Account Abstraction Update [see this explainer]
d. EIP-2929: Gas cost increases for state access opcodes
e. EVM 384 Update
f. EIP-2315 Simple Subroutines for the EVM - EIP-1559 Update
Next call: October 2nd, 2020 14:00 UTC
Does anyone notice the ASIC risky for ethash?
INNOSILICON offers new A10pro+ miner which has params: 720MH / 6G graphics memory / 1300w, and it will enter the market in Dec. with around 20T total hashrate at least.
So I am worry about the process of EIP-1057 and the chance for activating ProgPOW oneday
We didn't get to EIP-1057 last week, so it remains on the agenda. ethereum/EIPs#2894
I'm not sure -- perhaps @AndreaLanfranchi does -- but I think those units will not long survive DAG growth.
I think I'm wrong @haihongS. Those units might last on Ethash into 2024. https://investoon.com/tools/dag_size
yeah, very sad to hear this, @gcolvin -- and we don't expect too much about the schedule of ETH2.0 :)
We didn't get to EIP-1057 last week, so it remains on the agenda. ethereum/EIPs#2894
@Souptacular can we please get this added to the agenda
We didn't get to EIP-1057 last week, so it remains on the agenda. ethereum/EIPs#2894
This is a good idea except when, not if Ethereum is attacked prior to POS switch over, it will be so big and devistating that flipping a switch will already be too late. With the current public miners (we dont see the non public ones that are exteremely efficient) they will soon have 60% of the hash going into the 4GB drop off. This would be the best time IMO to attack the network. Is probably why Bitmain has not released a new miner.
@Souptacular I'd love to get some schedule time to talk about how Celo can contribute to ethereum/EIPs#2537 and ethereum/EIPs#2539
We've put together a bit of an explainer for account abstraction (EIP-2938) if anyone wants a bit more background before the call.
If there is time I would like to give an update on evm384.
@prestwich @SamWilsn @axic added to the agenda.
@axic: Do you have a URL to 384 we can use?
@Souptacular this can also be found on google, so I think it's safe to reference without explicit permission from Alex: https://notes.ethereum.org/@axic/evm384
Update: was the wrong document, sorry @axic, see comment below.
(cc @axic on googling this there is actually another version with a -preview postfix as top result which seems to be outdated, maybe it's worth to remove one of the documents to not have two concurrent versions floating around)
Also some...
Yolo v2 Update for EthereumJS
(can't join the call today). We're done with some necessary technical preparation to separate EIP and hardfork logic in the EthereumJS VM (essentially the VM can now also activate dedicated EIPs and not just a dedicated hardfork as some fixed set of EIPs). This base work provides the ground for a Yolo v2 integration. A big VM v5 release is planned for within the next 2-4 weeks.
Here is our Yolo v2 tracking issue, EIP-2315 (subroutines) and EIP-2537 (BLS precompiles) are ready and merged, EIP-2929 (state access gas cost increases) is currently in the works. All EIPs are planned to ship with v5.
We are also working hard to integrate the VM into our lately resumed client implementation, so that we can hopefully join the testnet directly within a couple of weeks. Don't take this as an official announcement yet though. 😄
@holgerd77 that is actually the update from last time :)
@Souptacular the new document is at https://notes.ethereum.org/@poemm/evm384-interface-update
EIP-2929 mentions an exception for calls to precompiles, but does not specify whether this exception is a current value or some new value, so I'd like to return EIP-2048 back to schedule for Berlin for consistency.
In don't understand what is unclear in EIP 2929, @shamatar, could you elaborate?
EIP-2929 mentions an exception for calls to precompiles, but does not specify whether this exception is a current value or some new value, so I'd like to return EIP-2048 back to schedule for Berlin for consistency.
In don't understand what is unclear in EIP 2929, @shamatar, could you elaborate?
It's may be a wording issue, but
Increase the gas cost of SLOAD (0x54) to 2100, and the *CALL opcode family (0xf1, f2, f4, fA), BALANCE 0x31 and the EXT* opcode family (0x3b, 0x3c, 0x3f) to 2600. Exempts (i) precompiles, and (ii) addresses and storage slots that have already been accessed in the same transaction.
reads as "keep the CALL cost to precompiles at 700 as before", whereas 2046 is about complete different deal - having STATICCALL to be ~100 for precompiles.
On a previous call it was mentioned that 2929 kind of supersedes 2046, but it doesn't look like this.
Link for ProgPow is out of date. Current PR is ethereum/EIPs#2894
@Souptacular
If there is time it would good to discuss which version of EIP-2315 we are moving forward with so far as restricting jumps. Most recent discussion here:
https://ethereum-magicians.org/t/eip-2315-simple-subroutines-for-the-evm-analysis/4229/54?u=gcolvin
@MadeofTin @Souptacular
Can we add breakout-room -1 and breakout-room -2 updates to EIP and upgrades updates, in case anyone missed it?
Also, a link to the tracker with clients update for YOLOv2 https://docs.google.com/spreadsheets/d/1BomvS0hjc88eTfx1b8Ufa6KYS3vMEb2c8TQ5HJWx2lc/edit#gid=408811124
If there is time it would good to discuss which version of EIP-2315 we are moving forward with so far as restricting jumps. Most recent discussion here:
https://ethereum-magicians.org/t/eip-2315-simple-subroutines-for-the-evm-analysis/4229/54?u=gcolvin
@MadeofTin @Souptacular
We should bump this up as it is related to yolov2 and therefore Berlin. @Souptacular
@shamatar that's just the abstracts. In practice, calls to precompiles (or already accesed addresses) costs 100
@shamatar that's just the abstracts. In practice, calls to precompiles (or already accesed addresses) costs
100
Do you plan to update an EIP as it's already merged is kind of a "source of truth"?
Increase the gas cost of SLOAD to 2100, and the CALL opcode family to 2600. Exempts (i) precompiles, and (ii) addresses and storage slots that have already been accessed in the same transaction.
You mean clarify that the exemption cost is 100 instead? The EIP can still be updated with clarifications, sure.
Increase the gas cost of SLOAD to 2100, and the CALL opcode family to 2600. Exempts (i) precompiles, and (ii) addresses and storage slots that have already been accessed in the same transaction.
You mean clarify that the exemption cost is
100instead? The EIP can still be updated with clarifications, sure.
Yes, that's what I've meant.
@gcolvin can you also link directly to the canonical specification or the current proposed versoin of progpow? There are just so many places of information, and hard to know which have bitrotted and which have not
@holiman The most recent PR should be merged today. Then we will have a canonical copy.
... and which one is that?