EIP-4844 Implementers Call #25

Transcript

okay uh welcome to 4844 call number 25. um so I guess the main things chat about today are just definitely um I don't think there were any significant spec updates um enough if I'm wrong this is the time to bring it up but uh nothing as far as I could see yeah the the we did release consensus specs last week that have the entire Fork build but that's not the target for devnet7 as I think discussed and we'll

talk about it again on the conflict this is player call but if your work if you're sincerely looking at that for the devnet7 you should still have the previous release but you should probably start parallelizing building out the other three features that are going into denim right actually yeah this is and then Alex has a comment as well about the pre-compile address so just I guess to reiterate uh so devnet 7 we agreed on all core devs it would be um basically just a reboot of devnet6 with uh hopefully the hive Test passing and I'm yeah and so this would not include the pre-compile address change

uh yes six is just dead uh I don't know if it's that right now does anyone have an update so six is kind of finalizing but with like an unfair distribution of keys it's like no I know there's a lot of work that people finalize Perry's manually signing that's what I've been doing part of us had this uh six pack I don't have edit access I don't know Perry If you do but maybe we should just rename this doc to say devnet six plus

seven and I think all the stuff we had that's PR candidates for devnet 7 should probably be candidates for devnet 8. and the one thing we should add alongside that is the pre-compile address change and are you implying we are likely to do another for for devnet for devnet8 or it might be full feature but it would probably have those it would definitely have the pr candidates in there as well I'm not sure yeah curious to hear from client teams but it felt like devnet6 didn't go super well and so I

don't know if we want to have like yeah devnet 7B just a rerunner um also have the Note 8 just focus on 444 I would I would like definite seven to be just a rerun yeah definite Aid I don't know kind of depends on how different uh seven goes in my opinion I'd say we keep doing this until we have like four eight four four nil down yep yeah I think I I agree introducing other complexity at this point seems crazy I like say I don't know yeah uh yeah

hello guys um I just wanted to say that probably to compile address changes uh are not so complex that could be included through uh it will allow a Q8 to reactant to update tests with new addresses and so on it might be good to have to uh this update uh I mean the compile update and probably uh update for uh given G2 operations which in conflict with this uh the computer address oh yeah what's the compiler picture what

do you mean about that specifically the the oh okay yeah yeah okay yeah that's right yeah okay so yeah pre-compile update should be in definite eight basically uh Enrico yeah I just want to say that um in Tech we are we are working directly on on Master Branch so we are kind of okay to relaunch then definite seven with the same characteristic of definite six uh because actually if there is something to fix we could Branch from previous commits if really want to do some hot

fix but that end of the net eight we definitely would like to have uh per alignment in what is inside the the current uh the nap merge spec otherwise we we are too too much uh this aligned with the Master's back because we want to start to implementing stuff and our the master Branch will not support another run of devnet 8 with an online spec so I guess fortunately 1744 and 70 45 are not crosslayer if we got to the point where um the consensus layer had seven had

Beacon root whatever that four seven eight eight uh implemented it could it could still be a no op on the execution layer because it's the only thing that that the IP does is pass it in so we do have some optionality in if people are in the position of wanting to get those couple features in even if the execution layer is not in accordance um so I guess we'll just kind of have to keep talking about it but it's good to know that that this is important to y'all I guess yeah uh before we move okay so I couldn't rename it so I just

copied over the contents okay got it uh okay so let's see how devnet7 goes this week I assume I guess is this a realistic timeline to launch it sometime in the next week or so I think so it looks like cats and other men are already passing tests I just haven't had a chance to run it against Aragon and bezub uh but if they do as well then we're good to go um their last message from Mario was that there's no CL engine API that's yet it's only El engine API test so

I guess we'll have to win the CL side got it and then one other thing that came up uh I don't know if this came up on Awkward answer where it bubbled up from but um there's a conversation so we discussed having basically Json IPC support for what was devnet7 and now would probably be devnet8 and one thing that came up as part of that was um making sure the like apis generally work well for the users of blob So like l2s um so I don't know uh Matt's not on the call but he brought this up um

but I guess that's something that would be good to do in the next couple weeks um just thinking through like is this art like do we have a Json RPC scheme that's like useful for ltus to uh submit blob and uh sort of interact with 4844 um and I guess Roberto you know from the maybe that's something that like the base team uh could like share a thousand because all the users like what would the correct apis look like uh for you to interact with this and yeah we in parallel we can reach out potentially some other l2s to get some

feedback yeah that's something we can try to provide and put on I don't think we thought it through very far yet but we can certainly spend some time on it right yeah I I don't think anyone has so I think it's probably as we get the consensus side of things stabilized that's probably the next thing we want so that um yeah we make sure that a lot makes sense yeah that's pretty much all we had it on the agenda I don't know if there's anything else people wanted to chat about um so on the lighthouse side we've had a couple of pain points with the ckcg lib

um so I was just wondering about a couple of open PR's there one of which is to allow us to compile the main net and minimal versions of the library and like using both at the same time which is that's really useful in our testing um so yeah it's I guess I was wondering the status if anyone has any stats on that or the likelihood of the pr I just Linked In the check it emerged yeah I don't think George is on the call

and you might know so I could try to find him offline and that was the biggest that was the biggest blocker you mentioned a couple PRS but is that one the biggest blocker for Sigma let's see this is the biggest one there's also another open PR about like the ckcg with Windows um which we're also gonna have to start out at some point okay uh yeah it doesn't seem like there's much many people from kcg on the call but we can definitely like this to them

um yeah I think anything else yeah I guess you want to know go ahead okay I would want to know uh when can we expect does a real trusted setup to be baked in from the ceremony so I know uh Carl and China are working on a blog post this week that will give

information about The Trusted ceremony wrapping up and the date that it will wrap up um so then we'll we'll have it at that point I don't know if that's you know plus two weeks or plus six weeks from now but it is certainly in the in the process it's shifting into the ending process so should be some information soon you're saying something as well um yes so if you're passing hard today

your your transaction pool implementation is actually wrong or at least has some has some issues uh there are two bugs in the in the hive tests right now um one is in the suit that does the transaction [Music] uh the transaction replacement so in this uh in this suit you send a few transactions and you can re like you replace other

transactions uh the problem is that only the gas price is increased not the data blob data gas price so that opens up a few attacks that you can do by replacing block transactions um and the other tests that is broken is the one that the multiple like the the transaction propagation one and this test also the gas price on gas price needs to be adjusted basically the

idea behind this test is you send five transactions with five blobs each to one node and five transactions with one blob each to the other node and then you should be then you should produce five five blocks with six six plops each so one transaction from each node uh the problem is that the transactions are priced the same and uh so it could be that you're building one like a block with one or two or five one block transactions instead of doing this like five plus one block so the the test is going to be

uh changed but that's not really an issue in the implementations because the transactions are priced the same so they should be okay to be put in but like clients might non-deterministically fail this this uh this check as a Geth did because there was no order between there's no basically no order between those transactions that are priced the same um so yeah but the the replacement one that is something that clients should look at into the in in their

uh transaction pool implementation it's fine for you if you want to allow data gas replacement transactions with the same data gas but I think it's bad because it can introduce a quite a lot of churn that's it thanks anyone my thoughts comments I think I can add some more context to why the test was written like that so the idea was that um since the data gas price doesn't really affect the miner or the profits in any way that the money

receipt could receive so the idea was that the data gas can be the same and it wouldn't matter to the to the miner um but yes as Mary said I guess it depends on on your the logic that you're applying to the to the replacement so I think I think it should discuss I can I can just change the test if that's okay but I think it should be discussed uh I don't regards what to do or which transactions to replace um yeah I guess yeah anything else just in

general around 4844 okay I'm yeah I guess it makes sense to keep probably a couple more of these calls um but and and see how they go but I think if um if they become shorter and shorter and we obviously cover a lot of this stuff in uh awkward devs uh maybe I don't know in the next month or so we can start phasing them out probably makes sense to do that like once we start moving from uh 4844 exclusive devnets to to like full dencoon devnets

but yeah I would I would at least keep the one uh two weeks from now any final comments for wrap up am I running that call yeah you are in two weeks uh okay yeah yeah and then if we do a really good job I won't have to run one in four weeks um so let's see how it goes cool cool thanks everyone I'll talk to you soon thanks bye thank you bye yeah see you guys