All Core Devs - Consensus #115

Transcript

foreign hey guys um yeah so today I will be covering the um old query death course this is uh meeting one one five uh copy and paste in the agenda here

Terence Tsao: let's see okay so first on the list is um death net eight um status so as far as we know definite 7 is um has been torn down so definite a is the next one so um from the last meeting even on Mondays um the net testing code it sounds like it was blocked by four seven eight eight but um but then we made a decision to

not click block by 478 so I guess in terms of like client status like um when are we comfortable with with a devnet 8 start date um yeah update here

oh

Parithosh Jayanthi: so maybe we can just start from uh um yeah so I've posted a link for Hive um I think most of the uh can we mute Maps yeah uh so from uh from the high perspective I don't know if Mario's here uh he isn't so okay I can just cover it

for now there were a couple of config issues those have all been fixed now and um you can see that most of the tests are indeed passing so so Barnabas and I are now just working on getting uh on updating our local testing tools and we'll just post a link on how different client teams can just make sure that there's no uh inconsistencies and in the meantime if if the client teams themselves are feeling confident then we're happy to start planning it and launch it early next week

Sanskrit

Terence Tsao: um does any in client team have an objection with the timeline so that is early next week

all right sounds great sounds like that's easy we'll do definite eight early next week and uh partner just ask our clan somewhere ready I guess I can speak for prison yep we are ready but yeah um I guess the question is um is any Clan that's not ready for early next week

Andrew Ashikhmin: with Aragon we are still working on um fixing the hive tests so we might be ready but it's still it's not certain

Terence Tsao: okay thanks for the update

so

Barnabas Busa: we have an image that we could use

already for ergon

no

Andrew Ashikhmin: no not yet we but it will be our in our demo in our default Branch uh but it's it's like no it's still um it's not working yet but next week it might be but yeah just I don't I don't know for sure we will we'll make another certainly

Terence Tsao: okay sounds good um are there anything else we want to cover for devnet a before we move on okay I got a thumbs up okay so the next topic is four seven a a so from my

understanding um this is there's a debate going on between regular contract versus system contract and the discussion is mostly on the El side so the CL side is fairly simple and easy to reason so um I guess um the decision like which way to go will be made in the next ACD Eco so it doesn't make sense to make the decision here but is there anything that people want to say to push towards further consensus okay teams say that we already decided for the salon contract yeah I I I guess

I'm just going through the agenda that like sounds like we don't want to make a decision here because um because because most of the El are not I guess some of the ER not here so it Sim size better to make a decision in the

Nets ACD eco

Tim Beiko: yeah can you hear me

Terence Tsao: yeah

Tim Beiko: okay yeah so we did decide to use a Smart contract rather than the pre-compile um and lifetime has a PR for that the thing we still need input from the El teams for and that it would be good to decide by the next acde is how we deploy that contract uh whether we just deploy it with like a regular transaction

before the fork or if we couple it with the fork so that during the fork activation we also do a contract deployment um that seems to be the biggest question and then as for the actual contract code like client uh PR has has that first pass at it um if people want to review that uh and you know potentially suggest improvements or or optimizations um we're also starting to reach out to Auditors the former who review the byte code um to make sure that it's um yeah it's it's

Terence Tsao: um thanks for the update team and Matt anything else you just want to cover here or any from anyone else

awesome it looks like the hype change already as well there is a hive issue a18 thank you Mario okay so next on the topic is the four Choice filter change for confirmation rule that is PR or issue three four three one so the toldr is that there is a new Fortress filtering change is two it's a simple two liner it's a refactor it has new tests so there is some debate on whether um to support changing for Choice logic at at the specific boundary as a part of the net release so we can group them as

close as possible versus do not enforce that so um so um so um so uh may make hell or anyone have any input on for this PR or issue

Mikhail Kalinin: um from the comments that are in this issue I can get through there is a slight preference and not to implement the strict working logic here and just make it make this change into the nap release of the clients

um so

Terence Tsao: right and even with the release we can assume most clients release will be grouped together we think like two to three weeks ish so yeah

Mikhail Kalinin: and the question is uh is there any concerns with this way of of making this change

Otis is not here

Terence Tsao: I was talking to poses and since I prison may have this change implemented already so the sub need to

double check within prison but yeah there's any other CL team have any concerns or input

Sean Anderson: um I'll just say from Lighthouse supporting the change at the epoch boundary is something we we would be able to do um if people do have preference for that as far as whether we should I personally would have to dig into the like change itself more to understand if there's any risk to not doing it

or

Ben Edgington: um teku I think already shared that we would uh have a slight preference for not forking at a specific Epoch but uh

releasing a version um prior to the um then up upgrade that supports the the new Fork Choice uh just because it it's a tiny reduction in complexity to do so um and um it just makes testing and all of that more straightforward having a sense for the risks that that would open up um there would be a very short window of time where potentially clients have there are multiple versions of the fault Choice out there um the circumstances in which that could cause trouble seem to me very remote but I don't know if anyone has a better sense for that than I do

Mikhail Kalinin: yeah I think about the probability that

something happens on the main map that will uh go the Streets between the clients that have this upgrade and between those that don't have it this upgrade it's really low um so it's more in Edge case very much even more edgy on the mainnet so yeah if if there is the um preference of the client team so not to activate this on eBook boundary then this is probably the way to go

Terence Tsao: so some side there's no strong objection

that we can move forward and um is there like a PR that to merge based on this I said 3431 and there's also tests that grow with it so would you say the next setup plan is essentially to approve and merge this PR 3431

Mikhail Kalinin: there is the comment from Lion which is sounds reasonable to me that we should merge this um yeah when the main main Network for the knob is that

we like merge the spr and uh and create test factors for it

Terence Tsao: okay so since I we are blocked by the Danette Harford airport and anyway but we do have rough consensus here that the this is what we prefer okay sounds good

and uh next topic is uh three four seven two which is uh clear we want to clarify aggregation Behavior it looks like currently clients do things slightly different like for example low star aggregate attestation as attestation comes versus prism it

Aggregates at 6.5 seconds so um the so the um so the motivation of this issue and the spec is so that client could be more consistent um it's pop here do you want to give like a high level intro for the pr

uh yes uh the motivation is that I noticed that uh system educator educate the SD station that alive only before 6.5 seconds so uh for the extension that Allied between 6.5 and 8 seconds you'll be not included in the in the PCM APK so and and that is why we are and and this VI is is to

uh specify specifically that the high-end should before a second and yeah but but but uh in my opinion uh the client should not like change like the high engine not work too much on on implementation to be consistent because like I I don't think it's worth a spending time but but I think it's it's at least uh a good idea to to be specific in in the specification yeah

yep

Terence Tsao: I tend to agree and I read through the issue it sounds to me there's just a couple points on like what language you

should use just like currently validated spec doesn't really use that must and shoot right and it sounds like it would be nice to start using those keywords more often and then at the end it sounds like we're agreeing that like must and must meaning that you must aggregate at a second and then you must send it's kind of impossible because there's always DeLay So Sansa is between like shoot versus shoot versus something like that

yeah that makes sense um there's

Terence Tsao: any client teams have any inputs for this PR yeah if there's no input I think we can

um it sounds to me yeah there is some sort of agreement here so we can just take you to the pr and I don't think Poltergeist is here POTUS is the one that's been voicing but yeah it sounds to me if today we can use the shoe keyword not the mouse keyword we should be able to merge this PR but yeah thank you Paul for uh opening the pr this is great so the next one is uh four Choice proposal booths in corporation so this one modify the unblock so it so it Awards The Proposal boost to the first plot rather than the last blog so this was one part of the post-mortem for the

RPC of bundling attack that was described in micros post and such that proposal boost is only set once per slot and then from the pr it seems like this is safe to Raw gradually so that's nice and um yeah that's an inclined team have any input on this 3352 and the um I believe I've seen a few approval already so sometimes there is no strong disagreement there but yeah does anyone have any voice or input on this one if there's no input I think um we can

merge this by end of this week shall we I mean yeah I think that's what we're looking for yep

cool all right I'm

Terence Tsao: going to the agenda again looks like if we covered everything yeah oh Perry wants to talk about um two million validator tests yeah Harry go for it

Parithosh Jayanthi: yeah hey so the last week's uh last week week and a half we've been running these um big validator test Maps um we've had I think three big a three relatively yeah there could be big attempts the

first two one that's successful uh we can change test two was the one we had last week um there's there's a detailed analysis posted over here the I can send you guys a link to the logs as well as some profiles that we've taken from last week as well but essentially at the 2.1 million Mark we weren't able to finalize um the theory at the moment is that uh there were too many validated duties some blocks coming in late leading to um not enough attestations being propagated throughout the network so even though we were seeing something like 80 percent uh proposals we were only seeing between 40 and 45 percent of attestations on the network so we

dropped on the uh overall valid data set size to 1.4 million now and we also dropped the number of keys per node to 3300. and we had the test the test and Genesis earlier today and this definite Genesis looks successful we're seeing relatively healthy rates of participation so hovering around the range of um 80 to 90 percent proposals and similarly 80 to 83 84 attestations so hopefully this test net is a bit more useful for client teams in order to run profiles on we leave it up until tomorrow morning so it would be great if

um if you guys can have a look if you can collect whatever profiles you need to and so on and we'd officially be making a proposal to have a 1.4 million validator set um public test net for holski and if there's any reason we shouldn't be doing that and we shouldn't be creating the chances state with that size then we'd love to hear about it now or over the next week of course

Terence Tsao: great thanks Perry um anyone have any um updates or feedback questions for Perry I guess I have a question whether like

is a 2 million still the goal that we're working in the background or are we like satisfied with 1.4 million right now

Parithosh Jayanthi: uh at the moment I think I would just go with 1.4 million um I think 2 million would recover would require a decent bit of work from a lot of clients and we're not necessarily sure if um if you want to have that in a public test Network or needing it will be a lot harder

rather make sense also will

Parithosh Jayanthi: have enough ether in the 1.4 million um Genesis states to grow it to 2 million eventually so we just buy more

time into it eventually if we need to

Terence Tsao: yep that makes sense and you would mentioned that the so the so the proposal is about 90 success rate at the station is about 80 to 83 percent um so the clients that's amazing the attestations um do they also know that

Parithosh Jayanthi: yeah I don't think anyone's a new single client is missing all the attestations I think it's more of uh every client isn't able to do 100 of the

attestations they should

Terence Tsao: I see cool that's great yeah I would encourage

people to take your look and this is happening on the Discord interrupt channel right

Parithosh Jayanthi: yeah that's right and everyone who's used to it has access the IP addresses are in the same places you always find them

Terence Tsao: great yeah thank you Perry for the update very useful and thank you partner they just posted the um girlfriend I think as well that's also very useful thank you guys anyone have any inputs for the test name or the testing status cool so that is it from the agendas perspective does anyone have like um any open discussion closing remarks

great this may be the shortest Accord thank you for making my job very easy yeah we were talking two weeks foreign thanks parents thank you guys bye

Speaker names adapted from ethereum/pm, CC BY-SA 3.0.