holesky coordination call transcript

holesky-0.txt

00:29 (Jim McDonald)
  Hello.

00:31 (Jim McDonald)
  Hi guys.

00:35 (Afri)
  Hey, how's it going?

00:42 (Jim McDonald)
  All good.

00:43 (Barnabas Busa)
  Hopefully we get to see people showing up here.
  Thank you for organizing it, by the way.

00:54 (Afri)
  Yeah, thanks for joining.

01:37 (Jim McDonald)
  Hi everyone.
  Hello.

01:48 (Rémy Roy)
  Hi everyone.

01:52 (Afri)
  Hello.
  How's it going?

01:56 (Rémy Roy)
  All fine so far.

01:58 (Barnabas Busa)
  On your side?

02:02 (Afri)
  Yes.

02:17 (Afri)
  Happy to see you are all right on time.

02:24 (Afri)
  Let's wait another two or 3 minutes for everyone to settle you.

03:08 (Jim McDonald)
  Hey, Remy.

03:51 (Afri)
  Hello everyone.

03:52 (Thorsten Behrens)
  It morning.

03:55 (Afri)
  Giving it one more minute because people are still joining.

04:04 (Jim McDonald)
  Hello.

04:46 (Afri)
  I guess I yet started slowly while.

04:48 (Afri)
  People are still flowing in.

04:50 (Afri)
  I'm happy that testnets is such a.

04:54 (Afri)
  Relevant topic that we have almost 40 people here.

04:57 (Afri)
  So welcome everyone.
  I'm afri I am with ChainSafe currently.

05:05 (Afri)
  I used to launch the Girly testnet.

05:08 (Afri)
  Back in the day.
  Did a couple of recaps and postmoderns.
  And Girly is currently in a very.

05:17 (Afri)
  Bad situation due to the total supply.

05:20 (Afri)
  Of tokens and due to the fact that it's being traded both on chain but also OTC.

05:27 (Afri)
  And.

05:30 (Afri)
  We discussed a couple of strategies for dealing with testnets going forward.

05:36 (Afri)
  And one of the strategies is to have regular replacement testnets.

05:41 (Afri)
  And today we want to discuss holeski.

05:44 (Afri)
  Testnet that is going to replace Girly hopefully this year or next year.

05:54 (Afri)
  For the meeting, I created a very thin agenda.
  Let's see how much time we have.

05:59 (Afri)
  First of all, I want to discuss the timeline.

06:01 (Afri)
  So if we want to launch at.

06:04 (Afri)
  A certain day in, for example, three months, then we need to define now.

06:08 (Afri)
  Which milestones we need to define.

06:11 (Afri)
  We need to define the milestones and also make sure we hit them in due time so that we have enough time for client teams to bake in the final Genesis, inspect and make releases accordingly.

06:24 (Afri)
  Then we need to discuss the initial validator set.

06:28 (Afri)
  So on Guinness, we will have a.

06:30 (Afri)
  Certain amount of validators that kick off the network.

06:33 (Afri)
  And we need to be sure this.

06:35 (Afri)
  Is a reliable validator set.
  And it also needs to be of.

06:38 (Afri)
  Certain size so that we can continue replacing early permanently.

06:44 (Afri)
  And then there are topics of lower.

06:48 (Afri)
  Priority, time permitting that would be discussing against location seeding developer accounts, strategies to distribute Easter in general, and so on.

07:04 (Afri)
  Pari just posted 5 minutes ago that.

07:06 (Afri)
  He also created a planning document.

07:09 (Afri)
  It's linked on GitHub.

07:11 (Afri)
  I just opened it like 1 minute ago.

07:13 (Afri)
  He added the item, defined the purpose of the test net.

07:16 (Afri)
  Makes sense, how big should it be, when do we do Genesis, who runs the keys?

07:21 (Afri)
  And so on.

07:22 (Afri)
  It's pretty much aligned with what I have in mind.
  So, yeah, let's kick this off.
  Does anyone have any general questions or.

07:30 (Afri)
  Does anyone need context before we start?

07:34 (Thorsten Behrens)
  I think I'm good.
  I'm looking at Paris notes right here.
  They make a lot of sense to me.
  So let's go through that.

07:41 (Afri)
  Yeah, sounds good.

07:44 (Barnabas Busa)
  I can also just walk everyone through this document because we wrote it together basically.
  So I'm Barnabas from the EF DevOps team and we are going to try to organize the launch of this testnet probably.
  So we are hoping to agree on a certain validator set that is the current main net.
  And this is very important because we currently don't have any testnets where we.

08:17 (Jim McDonald)
  Have more validators running than on the main net.

08:21 (Barnabas Busa)
  So in order to do that we.

08:23 (Barnabas Busa)
  Are thinking about 1 million to 1.5 million validators.

08:29 (Barnabas Busa)
  But this would be really up to the node operators.

08:34 (Jim McDonald)
  So either the client teams or DVTs.

08:38 (Barnabas Busa)
  Or other large node operators to agree.

08:43 (Jim McDonald)
  How many they want.

08:46 (Thorsten Behrens)
  Right.
  And just to be sure so what we're proposing here is a test net that has a large amount of validators by client teams and node ops and EF, et cetera.
  But is permissionless.
  So a true replacement for Gurley for people who want to just test.

09:06 (Jim McDonald)
  To.

09:06 (Barnabas Busa)
  Reach at least 66% by known entities.

09:12 (Jim McDonald)
  Basically trusted entities in order to get finality all the time would mean we.

09:20 (Barnabas Busa)
  Would need to have at least 700.

09:23 (Jim McDonald)
  To 800,000 that are known entities.

09:26 (Thorsten Behrens)
  Is there a desire to test non finality situations on this or is that a separate test net?

09:36 (Jim McDonald)
  No, not really.
  We don't plan to do that.

09:40 (Rémy Roy)
  Okay.

09:41 (Afri)
  So the idea is, if I may fill in girly code, is the only.

09:45 (Afri)
  Test net that is meant to be.

09:47 (Afri)
  Exactly as main net and if we.

09:49 (Afri)
  Deprecate Burly or shut it down we don't have anything like that.

09:51 (Afri)
  So we should aim for a test.

09:56 (Afri)
  Net that is as stable as possible.

10:01 (Afri)
  Or adheres to the same requirements you.

10:04 (Afri)
  Would have towards the main net.

10:08 (Barnabas Busa)
  So the general was that each client.

10:11 (Barnabas Busa)
  Team would be running about 100,000 validator.

10:13 (Barnabas Busa)
  And I do understand that this is a large amount of investment that they.

10:18 (Jim McDonald)
  Would need to make.

10:19 (Barnabas Busa)
  We are calculating that it would be about 20 nodes that they would need to run.

10:24 (Barnabas Busa)
  If you can run about 5000 node so you can see in the bottom.

10:30 (Barnabas Busa)
  Of the document some napkin math depending whether they would be hosting it on.

10:34 (Barnabas Busa)
  The cloud or in physical machines.

10:39 (Barnabas Busa)
  Around.

10:40 (Jim McDonald)
  1000 per nook or 2000 per month for running them in the cloud.

10:46 (Rémy Roy)
  Yeah.

10:47 (Thorsten Behrens)
  So for reference, I'm currently running 10,000 on nosis, on essentially one machine, bare metal, a rented cloud machine with six vCPU and I think 64 git memory and NVMe and that works.
  But that obviously costs more than a knock.
  The knock you just buy once and then that's that and then you just got bandwidth.

11:15 (Rémy Roy)
  Right, right.

11:16 (Barnabas Busa)
  But it would be more it comes.

11:19 (Jim McDonald)
  To trying to manage that probably.

11:22 (Barnabas Busa)
  Yeah.

11:22 (Thorsten Behrens)
  And we can get into the weeds as to how exactly to do these things.
  I don't want to get into the weeds too much but this would be.

11:29 (Barnabas Busa)
  A big upfront investment that client teams.

11:32 (Barnabas Busa)
  Would need to make.

11:33 (Barnabas Busa)
  And this is why this meeting is.

11:35 (Barnabas Busa)
  Very important where for sure you can.

11:37 (Barnabas Busa)
  See what kind of requirements we would, but I would like to hear from the client teams, like what kind of.

11:46 (Barnabas Busa)
  Size they were thinking.

11:49 (Jim McDonald)
  Before we do that, I'd suggest that.

11:52 (Rémy Roy)
  The number of validators kind of makes.

11:54 (Jim McDonald)
  It a number in itself.
  We need something that's significantly bigger than mainnet.
  Main net is running about 600,000 active validators.
  It's got about another 90 odd thousand in the queue and that queue hasn't.

12:04 (Jim McDonald)
  Been going down significantly for the past little while.

12:07 (Jim McDonald)
  So I think that if you kind of run the numbers through, you get to something around 1.1-1.2 million active validators.
  And I think we're just going to.

12:15 (Jim McDonald)
  Have to accept that's the number.

12:18 (Jim McDonald)
  If the client teams are and say we can run 20,000 validators each, that won't help us.

12:22 (Jim McDonald)
  We need something that looks a bit like main net.

12:25 (Jim McDonald)
  So I think we're going to have to go with those numbers and then.

12:28 (Jim McDonald)
  Work backwards to find out how we can actually get enough people to run those.

12:34 (Jim McDonald)
  Agreed.

12:35 (Thorsten Behrens)
  1.5 million sounds good.

12:38 (Rémy Roy)
  So just one thing on that.

12:40 (Jim McDonald)
  Actually before I forget, I don't care if it's 1.2 1.3 1.4 whatever, but one of the other things we should do is we should actually start off with a plan of exiting a percentage of everyone's validators pretty much as soon as the network kicks off.
  And the rationale behind that is, again, on main net, we got something like 70,000 exited validators and again we want to be replicating that.
  So maybe we'll have 200,000 validators in the mix.

13:05 (Jim McDonald)
  That should get shut down pretty much.

13:07 (Jim McDonald)
  Immediately because that just makes the validator state information look a little bit more like main net or main net plus rather than making it all clean and pleasant.
  Because as we've seen before right, there's always little corner edge cases.
  So we want to replicate that as.

13:21 (Jim McDonald)
  Close as we can.

13:27 (Barnabas Busa)
  Yeah, that should be no problem.

13:30 (Mario Havel)
  Regarding the requirements, I think that hardware itself is not such a problem.
  Like even like some second hand server can run hundreds of thousands of validators.
  But the issue I find is the bandwidth element.
  How is the bandwidth requirement growing with more validators?
  Does it get stable?
  Is it like logarithmic?

13:53 (Thorsten Behrens)
  So you're looking at about twelve terabytes a month once you're above 70 validators.
  I know that Sigma prime have been looking at ways of reducing that requirement.
  I'm not sure where that discussion is right now, whether that will make it into denip or whatever.
  The client teams will know that better than I do.

14:16 (Mario Havel)
  Yeah, thanks.
  Yeah, I have to look into that because with many validators, I think the bandwidth will be the main constraint.

14:23 (Thorsten Behrens)
  Yeah, it's really per node.
  Right?
  So you're looking at twelve to 14 terabytes a month per node.
  Roughly 50 incoming outgoing.
  Not exactly, but roughly.
  So once you put that into cloud like AWS or whatever, it gets stupidly expensive.
  If you put it into something that's a bare metal with a fixed budget, like some of these bare metal offerings are like you get 25 terabytes a month for your server done, right?
  That's sufficient, that's fine, that's not an issue.

15:02 (Mario Havel)
  Got it.

15:03 (Afri)
  Thank you.

15:03 (Mario Havel)
  Yeah, I'm going to run it physically, actually from Holoshovitz themselves and yeah, the thing is just like you have enough bandwidth left for other stuff so got to upgrade the connection.

15:18 (Barnabas Busa)
  Large amount of Validators will consume about.

15:21 (Jim McDonald)
  100 megabits per second, I would say.

15:24 (Barnabas Busa)
  So if you have anything more than.

15:26 (Barnabas Busa)
  That you should be good to go.

15:31 (Barnabas Busa)
  Also if you exposing your ports for.

15:35 (Mario Havel)
  PTP communication, there is also other infrastructure running and it's 300 megabyte connection but it already consumes like 150 of that.
  So I'm going to probably upgrade to 500 and run like a bunch of Validators.
  Yeah, looking forward to that.

16:00 (Barnabas Busa)
  Yeah.

16:01 (Thorsten Behrens)
  So funding is a good question.
  Right?
  So if I think about my own operations, I like testnets, I'm perfectly happy to run things.
  And in bot I don't have a data center.
  This would have to run in basically rented servers which are about $160 per server per month.
  So I don't know how that is envisioned right now.

16:30 (Barnabas Busa)
  So the vision would be that primary.

16:33 (Barnabas Busa)
  Target would be declined teams.
  The secondary market would be the node operators like Lido and other large operators.

16:42 (Barnabas Busa)
  And then everyone else goes after that.

16:47 (Barnabas Busa)
  So we would like to keep that to the minimum.
  We would only want to fund it if really there's nothing coming together.

16:58 (Rémy Roy)
  Makes a lot of sense.

17:01 (Barnabas Busa)
  Also just for decline plan, ideally we.

17:05 (Barnabas Busa)
  Would like to run this for at least five years.
  We're going to be deprecating it after five years.

17:12 (Jim McDonald)
  So just make your plans according to that.

17:17 (Jim McDonald)
  So on that I would be much.

17:19 (Jim McDonald)
  More inclined to say three years for various reasons.

17:22 (Jim McDonald)
  One is five years is a serious commitment, it might put people off.

17:25 (Rémy Roy)
  Second is that depreciation schedules commonly run.

17:29 (Jim McDonald)
  On three years so people can amortize.

17:31 (Jim McDonald)
  Our hardware correctly over a three year schedule.
  And finally, we shouldn't be leaving networks running if they're test nets for that length of time.

17:40 (Jim McDonald)
  I mean, we're going to see this with gurley, right?

17:42 (Jim McDonald)
  There's a whole bunch of infrastructure that's deployed on there has been deployed once.

17:46 (Jim McDonald)
  No doubt by hand.

17:48 (Jim McDonald)
  There are no methods or systems to deploy it.

17:50 (Jim McDonald)
  We've had Sepolia running for a while and a lot of people haven't moved.

17:53 (Jim McDonald)
  Over to it even though it's a far friendlier environment because all of the contracts are missing.

17:58 (Jim McDonald)
  So I think if we say you.

18:00 (Rémy Roy)
  Have something to do now and you.

18:03 (Jim McDonald)
  Get to do it again in five years, they'll just do something now and forget about it.
  Maybe if we make it clear that.

18:07 (Jim McDonald)
  It'S a relatively short time scale, something like three years, they'll start to make.

18:11 (Rémy Roy)
  Sure that they have the processes to.

18:13 (Jim McDonald)
  Actually deploy base contracts to a new chain ENS and all the DEXes and.

18:19 (Jim McDonald)
  All this kind of stuff.
  Because of that I'd be inclined to keep it actually on a faster period and something like three years rather than five.

18:28 (Rémy Roy)
  But that's just my view.
  I also agree with a shorter period.

18:43 (lightclient)
  The problem is you have to think about it from the perspective of the other infrastructure providers.
  Things like scan things like infuria they have a lot of fixed upfront costs in application developers, they have this fixed upfront cost of using testnets and switching to testnets is kind of an expensive thing for them.
  And so okay, yeah, sounds nice for the people who have to operate the test net.
  Sorry, I just got kicked out of the building.

19:16 (Afri)
  I would carefully say three years or.

19:19 (Afri)
  Five years, it is a difference but.

19:21 (Afri)
  It'S a bit out of scope for today's call.
  We had this is like a very involved discussion.

19:27 (Afri)
  I've been dealing with this for a.

19:29 (Afri)
  Couple of months now with application teams.

19:32 (Afri)
  With a two teams, with infrastructure providers.

19:35 (Afri)
  And there's so many different opinions on this.

19:37 (Afri)
  So I would say just keep in.

19:39 (Afri)
  Mind it might be three, four or.

19:41 (Afri)
  Five years but it's for this today's call not important.

19:44 (Afri)
  Just keep in mind this is infrastructure that need to be maintained long term.

19:51 (Afri)
  And maybe we can talk about timing.

19:59 (Barnabas Busa)
  The proposed date was the 15 September.
  Is that acceptable by all client teams?

20:13 (Jim McDonald)
  How to tell?

20:13 (Jim McDonald)
  Right for the moment?

20:14 (Rémy Roy)
  I mean one thing, I'd say a.

20:16 (Jim McDonald)
  Couple of things here, right?

20:18 (Jim McDonald)
  One is that I'm not sure launching.

20:20 (Jim McDonald)
  On any anniversary or anything is necessarily.

20:22 (Jim McDonald)
  The right way of doing it.

20:23 (Jim McDonald)
  But what I will say is actually my inclination would be certainly given some of them comments and the concerns that.

20:29 (Jim McDonald)
  Come up so far, we should look.

20:31 (Jim McDonald)
  At doing a soft launch.
  So datex doesn't matter when but you.

20:36 (Rémy Roy)
  Know, datex the network goes live, give.

20:39 (Jim McDonald)
  Everyone a week to stabilize their infrastructure and then check on the validator performance for a week.

20:46 (Jim McDonald)
  So what I mean by that is like we said, if potentially someone's running on an environment where they haven't got the bandwidth or they said they were going to run validators and haven't whatever.

20:54 (Jim McDonald)
  It might be at the end of.

20:56 (Jim McDonald)
  That second week we can say okay.

20:58 (Jim McDonald)
  This worked, this didn't.
  Now ideally everyone will be running fine.

21:02 (Jim McDonald)
  Everything will be happy and then we.

21:04 (Jim McDonald)
  Can just announce that the testnet has launched and available.

21:08 (Jim McDonald)
  If it's not working, let's say we got 20% of the validators already that.

21:12 (Jim McDonald)
  Aren'T actually managing to validate.

21:14 (Jim McDonald)
  Then we shut it down.
  Regenesis without those validators and relaunch from.

21:21 (Jim McDonald)
  Having watched the Polio, which was a.

21:23 (Jim McDonald)
  Tiny network, there was certainly a significant.

21:27 (Jim McDonald)
  Amount of time where people didn't have.

21:29 (Jim McDonald)
  The systems up and running or working properly and or they tried and failed.

21:34 (Jim McDonald)
  To do it and or they lost their keys and or all sorts of other things.

21:37 (Jim McDonald)
  So for a network of this size, lots of people are going to sign up and say, yes, they want to do it.

21:42 (Rémy Roy)
  But I think that if we don't.

21:44 (Jim McDonald)
  Want to start with 70% participation, we should probably make sure that we do a soft launch, make sure that everyone.

21:53 (Jim McDonald)
  Who is involved are actually able to.

21:55 (Jim McDonald)
  Run the validators before we officially launch and, like we said, have that fallback.

21:59 (Jim McDonald)
  Of being able to shut down and regenesis if there's a significant percentage of.

22:04 (Jim McDonald)
  The validators who are failing to validate.

22:10 (Afri)
  Are you suggesting that we split between launching a network and announcing it?

22:17 (Jim McDonald)
  Yeah, basically.
  So basically for the first two weeks of the network, we're not guaranteeing it'll survive.

22:24 (Jim McDonald)
  So if the first two weeks to the launch, we'll say we're trying, it.

22:27 (Jim McDonald)
  Making seeing if the validator set is manageable by the people who are running the nodes, two weeks later, it's either, yeah, we're happy and so the thing is now going to stay around for three years or we're not happy and.

22:39 (Jim McDonald)
  We'Re going to relaunch.

22:43 (Afri)
  I think this is pretty sensible.
  So if we take if we set.

22:48 (Afri)
  A Guinness date on September 15 or.

22:51 (Afri)
  We would reevaluate around October 1 or even earlier.

23:00 (Jim McDonald)
  Yeah, I think that makes sense.
  Like we said, we probably need a.

23:03 (Jim McDonald)
  Week for initial stability.

23:05 (Jim McDonald)
  I don't think we've seen one test net yet where someone hasn't missed Genesis, so people are going to miss it or we're going to catch people on.

23:11 (Jim McDonald)
  The tail end of vacations or whatever else is happening.

23:15 (Jim McDonald)
  So give everyone a week to get the validators up and running, and then the next week we look at actually.

23:19 (Jim McDonald)
  The performance of the network and then.

23:21 (Jim McDonald)
  Like we said, if we're happy, launch.

23:22 (Jim McDonald)
  If not, then reevaluate.

23:27 (Afri)
  Okay, just taking notes.

23:30 (Afri)
  Here another question for client teams.

23:34 (Afri)
  If were to launch on September.

23:36 (Afri)
  15, what would be the deadline for Guinnesses and Config freeze?

23:41 (Afri)
  Would four weeks be enough or do.

23:42 (Afri)
  You need more time for releases from our site?
  Four weeks will be enough.
  Yeah, same for Lodestar.

23:57 (Afri)
  It doesn't take too much to implement.

24:00 (Afri)
  A new config for test net.

24:06 (Afri)
  Okay, in this case, unless anyone needs.

24:08 (Afri)
  More time, I would propose that we.

24:11 (Afri)
  Release the final Genesis state and config.

24:16 (Afri)
  On August 17, which is roughly four weeks.
  Or is it in five weeks?
  It's four weeks and one day before Genesis in around nine weeks.

24:46 (Barnabas Busa)
  The major question would be the validator counts per client team.
  Because if someone has to spin up 20 nodes, that might be something that takes them quite a bit, especially if.

24:57 (Barnabas Busa)
  They never run their own infrastructure before.

25:03 (Barnabas Busa)
  So that's why it would be very important to agree on the certain set of validators for each client and then.

25:11 (Barnabas Busa)
  We can go from there.

25:18 (Afri)
  Okay, but just for now, we don't.

25:21 (Afri)
  Have to finalize the timeline yet, but.

25:22 (Afri)
  Just assume we do this timeline, like.

25:25 (Afri)
  Put out release on our 17 and September 15.
  So we should walk backwards now and see which client team can commit to how many data.

25:36 (Afri)
  Does this make sense?

25:40 (Jim McDonald)
  So obviously there's a lot of people on this call.

25:44 (Rémy Roy)
  I would suggest you have two lists, okay?

25:47 (Jim McDonald)
  One is your list of the people.

25:48 (Jim McDonald)
  Who you want to be running the.

25:50 (Jim McDonald)
  Majority of the network.
  So they'll be the client teams, EF, whoever else you want on that list, okay, that list.

25:57 (Jim McDonald)
  Then you say, okay, we want to get maybe at Genesis you want maybe.

26:01 (Jim McDonald)
  80% of the validators to be on that list.

26:04 (Jim McDonald)
  Then you have another list of all.

26:05 (Jim McDonald)
  The other people who want to run their own nodes and run their own.

26:10 (Jim McDonald)
  Validators, and you put them on a second list.
  And then basically you apportion everything out according to what they say.
  So let's say Team A says, I'm going to run 100,000 validators and Team B says, I can run 50,000 validators.

26:23 (Jim McDonald)
  That's fine if it ends up that.

26:26 (Jim McDonald)
  You have more allocation than available.
  So let's say you've got the client teams between them sign up for a million ether and you only want to.

26:36 (Jim McDonald)
  Give them 750,000 or whatever validators.
  Excuse me.

26:39 (Jim McDonald)
  Then you just multiply everyone by zero.

26:41 (Jim McDonald)
  Point 75 and that's the number they get.

26:44 (Jim McDonald)
  Basically everyone sort of gets what they want proportionately.

26:47 (Jim McDonald)
  But everyone should put in a maximum, right?

26:49 (Jim McDonald)
  Because if someone signs up to do 20,000 validators and you give them 50, that's a problem.

26:54 (Jim McDonald)
  But if you can say that each.

26:58 (Jim McDonald)
  Client team signs up to what the.

26:59 (Jim McDonald)
  Maximum they're happy to do, then that should give you the ability to know.

27:04 (Jim McDonald)
  How much each one can have, right?
  And the other thing with that is, like we said, you don't have to get involved with saying all how much can X run and Y will run.

27:15 (Jim McDonald)
  And how much we got, right?
  You just get a maximum and then you work it out proportionately.

27:19 (Jim McDonald)
  That's type thing.
  That should be able to get done.

27:20 (Jim McDonald)
  In a week or so at most.

27:26 (Afri)
  I mean, we could even do it.

27:28 (Afri)
  On a call right now.

27:29 (Afri)
  If Team CEO can commit to a.

27:31 (Afri)
  Certain number or propose a certain number they are willing to commit to, it.

27:35 (Afri)
  Doesn'T need to be finalized today.

27:37 (Afri)
  We can take this Async and do.

27:40 (Afri)
  Another one or two weeks so we.

27:41 (Afri)
  Can check back with your infrastructure teams.

27:46 (Afri)
  But if everyone was on this call.

27:49 (Afri)
  Can commit to a number and I.

27:51 (Afri)
  Just have pen and paper here and just write down and we do like some math now.
  What do you think?

28:03 (Afri)
  Let me start.

28:04 (Afri)
  I just write down load star one hundred K.

28:10 (Afri)
  And I know you have DevOps also wants to run.

28:15 (Afri)
  100K.

28:17 (Afri)
  Anyone else who wants to commit right now?

28:21 (Jim McDonald)
  So just so you know, the average.

28:23 (Jim McDonald)
  Needs to be around 80,000 between all the client teams.
  Similar, if you want to get an.

28:29 (Jim McDonald)
  80% network run by the year and.

28:32 (Jim McDonald)
  Client teams.

28:42 (Matt Nelson)
  This is the basic team.

28:43 (Rémy Roy)
  We can probably commit to that number.

28:46 (Matt Nelson)
  I need to double check because we're running like a handful like maybe 5000 on girly so they just need to double check the kind of budget for that increase.
  80 to 100,000 sounds a reasonable number and we run other infrastructure elsewhere so I'm sure it won't be too much of a problem.
  Just want to double check.

29:10 (Barnabas Busa)
  Before we ask for hard commitments on this call right away.
  We might need to have another call in a week or in two weeks where we can actually have a hard.

29:21 (Barnabas Busa)
  Commitment from each client teams.

29:23 (Barnabas Busa)
  But this is a good place to have an agreement on what we are expecting and then maybe in two weeks.

29:30 (Barnabas Busa)
  We can have the hard agreements by.

29:32 (Barnabas Busa)
  The client teams and also some node operators.

29:37 (Jim McDonald)
  The other point is you probably want to decide which node operators you want to be running that kind of 80%.

29:43 (Jim McDonald)
  Of systems because obviously the more they.

29:45 (Jim McDonald)
  Run, the fewer that the client teams.

29:48 (Jim McDonald)
  Have to run and the cheaper it.

29:50 (Jim McDonald)
  Will be for them and the easier.

29:51 (Jim McDonald)
  It will be for them to commit.

29:55 (Jim McDonald)
  This is kind of why I thought it would work better offline.
  I think you're going to struggle to.

29:59 (Jim McDonald)
  Get hard commitments from people when they're.

30:02 (Jim McDonald)
  Signing up to paying tens of thousands.

30:04 (Jim McDonald)
  Of dollars or whatever over a lifetime.

30:06 (Jim McDonald)
  Right now maybe some can but.

30:21 (Afri)
  Then.

30:22 (Afri)
  We make it a homework.
  Everyone checks in for the maximum possible.

30:29 (Afri)
  Keys they can commit to and then.

30:31 (Afri)
  We just have a subsequent call where.

30:34 (Afri)
  We have goes through hard commitment.

30:51 (Jim McDonald)
  So an aside but just something that.

30:54 (Jim McDonald)
  Popped up on the chat was talking about keys.

30:59 (Rémy Roy)
  I would also probably guess that we.

31:02 (Jim McDonald)
  Should run a proportion if not all.

31:04 (Jim McDonald)
  Of the key, all the Validators created with BLS keys.

31:08 (Jim McDonald)
  Again, so there'll be a lot of.

31:09 (Jim McDonald)
  BLS two execution change addresses in the chain and in the state.

31:15 (Jim McDonald)
  So again if we start them all with version one credentials and there'll be no changes and obviously again it doesn't.

31:21 (Jim McDonald)
  Really map to how mainnet looks.

31:23 (Jim McDonald)
  So I'd suggest that when these are.

31:24 (Jim McDonald)
  Created, they are created with type zero credentials and people can set type ones as and when they desire.

31:33 (Jim McDonald)
  The other thing I would just say, and this is aimed squarely at Barnabas I suspect is that I suspect there's teams out there, DVT and SEMA that.

31:41 (Rémy Roy)
  Will want to create their own keys for Validators.

31:44 (Jim McDonald)
  So you may need to think about.

31:46 (Jim McDonald)
  How you want to handle that.
  That's more of an operational thing further.

31:49 (Jim McDonald)
  Down the line but we can't just hand out Mnemonics because some people for.

31:53 (Rémy Roy)
  Some systems won't work.

31:57 (Barnabas Busa)
  It's also possible that they can just.

31:59 (Barnabas Busa)
  Make deposits as soon as we launch Genesis, correct?

32:04 (Jim McDonald)
  Yeah, another really good point actually.

32:10 (Jim McDonald)
  To.

32:10 (Jim McDonald)
  Be clear we're talking about the Genesis Validator set but I think that we.

32:13 (Jim McDonald)
  Should make sure that there are a lot of.

32:19 (Jim McDonald)
  There'S a lot of ether.

32:20 (Jim McDonald)
  Floating around on this network.
  So for example, you can create all.

32:23 (Jim McDonald)
  The initial validators with say a million ether each in addition to, and I know after you talk about this, in addition to giving the devs and the users some initial allocation that just allows.

32:32 (Jim McDonald)
  People to create new validators as and when.

32:36 (Jim McDonald)
  So again, you know, this is a.

32:38 (Jim McDonald)
  Permissionless system, that list of validators is not final.

32:42 (Barnabas Busa)
  Right.

32:43 (Barnabas Busa)
  The idea was that we're going to.

32:45 (Barnabas Busa)
  Be inflating the supply just like as.

32:47 (Barnabas Busa)
  You said, the same way as we did for Sepolia.
  So as soon as someone makes a.

32:52 (Jim McDonald)
  Withdrawal they will have a lot more ease available.

32:56 (Rémy Roy)
  Sounds good.

32:58 (Jim McDonald)
  That's what the idea.

33:08 (Afri)
  Yeah.

33:09 (Afri)
  Allocation is a topic that is also.

33:11 (Afri)
  Important but this is something we can.

33:14 (Afri)
  Continue discussing async there were some good ideas around.

33:19 (Afri)
  We had this team that did the.

33:22 (Afri)
  Testnet token claim on Girly and Sepolia.

33:26 (Afri)
  And they had a pretty large database of addresses that had been active in the past.

33:33 (Afri)
  So we might be able to utilize this to do like an initial seat of.

33:40 (Afri)
  Accounts after Guinness.
  So we distribute initially a lot of.

33:45 (Afri)
  Ether on the network but also allocating.

33:49 (Afri)
  Enough ether for validators and additional validators.

33:56 (Afri)
  In future is also very sensible to think about it.

33:59 (Afri)
  What I would be careful about is.

34:01 (Afri)
  Everything that goes towards having a custom deposit contract, having custom withdrawals or inflating through withdrawals.

34:09 (Afri)
  Just maybe, but this is open for.

34:11 (Afri)
  Discussion, just maybe because we should be very pedantic to have at least one.

34:14 (Afri)
  Test net, one to one maps to main net.

34:21 (Jim McDonald)
  So this would be genesis state of.

34:24 (Jim McDonald)
  The consensus layer, right?

34:25 (Jim McDonald)
  So it wouldn't involve any changes to.

34:27 (Jim McDonald)
  Deposit contract or anything like that.

34:30 (Afri)
  Yes, if you allocate ether to validators.

34:34 (Afri)
  Right, but if you modify deposit contract.

34:36 (Afri)
  Then I would carefully say we should not do that.

34:41 (Jim McDonald)
  I agree with that.

34:42 (Jim McDonald)
  No, we shouldn't change a deposit contract.
  So again SAPOLI is permissioned.
  You need tokens to create validators.
  We have to have some environment somewhere.

34:52 (Jim McDonald)
  Where average new user can turn up and actually go through the deposit process to test.

34:57 (Jim McDonald)
  So I think this should be say.

34:59 (Jim McDonald)
  As vanilla as possible.

35:04 (Jim McDonald)
  Yes.

35:05 (Barnabas Busa)
  We don't plan to have a different token for deposits.

35:08 (Barnabas Busa)
  Just going to be the native token.

35:10 (Barnabas Busa)
  The native.
  We don't plan to change the deposit.

35:17 (Jim McDonald)
  Contract.

35:29 (Barnabas Busa)
  And we plan to launch with epoch zero and we're going to go through the map.

35:54 (Barnabas Busa)
  Are there any questions from anyone?

36:00 (Afri)
  Just try to follow the chat.

36:02 (Afri)
  I would appreciate if you can have.

36:03 (Afri)
  The conversation through voice so I can.

36:07 (Afri)
  Have to focus on two things.
  The question is if we launch the.

36:11 (Afri)
  Standup config.

36:15 (Barnabas Busa)
  We will not check 1 second.

36:27 (Afri)
  No problem.
  Hello.

36:34 (Matt Nelson)
  Thomas from Rockway X asked if.

36:38 (Rémy Roy)
  There'S any plan for a validator faucet eaveseeker would be happy to help with that.

36:49 (Barnabas Busa)
  Plenty of funds floating around so I'm.

36:52 (Jim McDonald)
  Sure it's going to be easy to get.

36:55 (Jim McDonald)
  Yes.

37:00 (Martin Vejmelka)
  Martin from Rockox will also be pretty happy to help with a Faucet.
  Like, we had in mind a very similar idea that was floated to test the original Genesis Validators.
  So I was wondering if it would be possible to sort of run a full node first for a period of time to show that you're able to actually operate some hardware.
  And then we could be scanning that.
  And if you could keep Good uptime for a few days or like we can decide on the parameters here, then the Faucet would be able to deposit you the third to Ethan.

37:31 (Afri)
  You could run your own Validator in.

37:35 (Martin Vejmelka)
  This way before getting the ETH.
  You actually demonstrate intent, but you also.

37:39 (Afri)
  Show competence in the target area.
  Does that make sense?

37:46 (Jim McDonald)
  I suspect there's going to be so.

37:47 (Jim McDonald)
  Much of this ETH floating around on this testnet that people will just spin up validators and they'll be forces to just spit out thousands of ETH and.

37:55 (Jim McDonald)
  All that kind of stuff.
  A lot of the plan here is.

37:58 (Jim McDonald)
  To avoid that issue with goalie, where suddenly it's scarce and hence worth money and hence hoarded.
  So I would very much hope if.

38:08 (Jim McDonald)
  We do this right, that we're never going to need gateways to obtain a.

38:14 (Jim McDonald)
  Reasonable amount of ether either for a.

38:16 (Rémy Roy)
  Validator or just for use.

38:19 (Martin Vejmelka)
  I hear that makes sense.

38:23 (Barnabas Busa)
  One question that might arises is the initial Genesis State that's going to be rather big.
  Are the client teams able to facilitate.

38:33 (Barnabas Busa)
  Very large Genesis State files or is.

38:38 (Barnabas Busa)
  This something that they would need to.

38:39 (Barnabas Busa)
  Work on or test out?

38:41 (Jim McDonald)
  Possibly.

39:00 (Barnabas Busa)
  I remember back in when I was testing with your wallet seven or six, we did 600,000 in the Genesis State.

39:09 (Barnabas Busa)
  And that was causing some issues with.

39:11 (Jim McDonald)
  Some of the client things.

39:13 (Barnabas Busa)
  I had to set some certain flags to fix it.

39:18 (Barnabas Busa)
  I'm just curious if this is something.

39:20 (Barnabas Busa)
  That will work with like 1.
  5 million validators.

39:24 (parithosh)
  So we should probably do a private test beforehand, just with few notes, make sure that all the clients are working, and then we can do the real.

39:33 (Jim McDonald)
  One on the day.

39:34 (Jim McDonald)
  Yeah, that also sounds good.

39:47 (Jim McDonald)
  Okay.

39:47 (Afri)
  Looking at the clock, I tried to take a look at my notes and.

39:50 (Afri)
  Just conclude what we discussed today.
  We're trying to launch around 1.5 million validators as possible.

40:01 (Afri)
  We try to have around 70% by known entities so that we can guarantee.

40:07 (Afri)
  Finality that needs to be covered by roughly mandated client teams and node operators.
  We discussed that this is on average around 880 to 100K validators per team.

40:26 (Afri)
  We also decided not to make hard.

40:28 (Afri)
  Commitments today because it's a lot of.

40:30 (Afri)
  Requirements for the team.
  So we give every team some time.

40:33 (Afri)
  To talk to their infrastructure teams and.

40:38 (Afri)
  Make sure this is feasible and then come back in a subsequent call and make hard commitments in the coming weeks.

40:49 (Afri)
  And the priority here would be number one.

40:52 (Afri)
  Priority would be client teams.
  Number two priority would be node operators.

40:56 (Afri)
  And everyone else gets into the third.

41:00 (Afri)
  Category and we want to make sure.

41:02 (Afri)
  That the first category covers at least 65% or 70%.

41:09 (Afri)
  We also discussed the timeline.

41:12 (Afri)
  So we want to try to stick.

41:14 (Afri)
  To the September 15 Genesis and we.

41:19 (Afri)
  Want to do release of Genesis state.

41:22 (Afri)
  And configuration on August 17 which is roughly four weeks before that.

41:27 (Afri)
  We will do a private test before.

41:28 (Afri)
  And just to make sure the state.

41:31 (Afri)
  That we agreed upon is actually launchable.

41:35 (Afri)
  And we also decide not to announce.

41:38 (Afri)
  The genesis or the launch of the testnet until we have finality guarantees by the network and by the teams running.

41:46 (Afri)
  So we want to make a subsequent decision around October 1 where we make.

41:50 (Afri)
  The decision to either announce it if it's working out well or to decide.

41:54 (Afri)
  For regenesses in case we need to start over.

42:00 (Afri)
  We will launch with Capella Shanghai configuration.

42:04 (Afri)
  On Genesis and we discussed some other.

42:08 (Afri)
  Items such as exiting some portion of.

42:12 (Afri)
  The validance right after Genesis to be.

42:13 (Afri)
  More like mainnet and also to have the majority of.

42:22 (Afri)
  Withdrawal keys with type zero credentials.

42:30 (Afri)
  Is there anything I missed or is.

42:32 (Afri)
  There anything we need to discuss today?
   In addition to that.

42:41 (Jim McDonald)
  Was there a.

42:42 (Jim McDonald)
  Note in there about the number of validators that we should exit immediately just to get them in state as well?

42:51 (Barnabas Busa)
  We can probably exit 200,000 yeah, right away.
  And also set up 200,000 in the.

43:00 (Barnabas Busa)
  Deposit queue so we might as well have a bit of everything.

43:09 (Afri)
  Sounds good.

43:12 (Afri)
  Then I would try to schedule a subsequent call.
  This call was a bit short noticed.
  Does time in general work for all.

43:21 (Afri)
  Of you or should we try to find a better time?

43:28 (Jim McDonald)
  We got 50 people here.

43:30 (Jim McDonald)
  It seems like it's a good enough time.

43:31 (Afri)
  Right, okay.

43:34 (Afri)
  So with regard to the hard commitments.

43:37 (Afri)
  I will be unfortunately I will be on vacation for the next three weeks.

43:40 (Afri)
  So if you want to schedule this.

43:42 (Afri)
  Intermediate call in the coming weeks and.

43:46 (Afri)
  Someone else has to step up otherwise.

43:48 (Afri)
  I would propose to do it more towards mid July in around four weeks.

43:55 (Jim McDonald)
  So I would say in the meantime.

43:57 (Jim McDonald)
  We should probably throw up three issues on the well pick your favorite repo where people can commit to how many validators they want to run.
  So as you mentioned, you got kind of three different sets of people that will allow you to gather the commitments.

44:13 (Jim McDonald)
  Asynchronously and allow us to actually work out what's available.

44:17 (Rémy Roy)
  Right.

44:17 (Jim McDonald)
  Because that will give everyone an idea.

44:20 (Jim McDonald)
  Of how much we can push out.

44:23 (Jim McDonald)
  There and if there's any holes like.

44:25 (Jim McDonald)
  Say for example we can't get to.

44:27 (Jim McDonald)
  60% of the network or whatever running by client teams then that's something we.

44:32 (Jim McDonald)
  Can look at then.

44:35 (Afri)
  Sounds good.

44:36 (Afri)
  We have the Holistic repository in the.

44:38 (Afri)
  Clients organization and I can just create.

44:40 (Afri)
  Three issues for the three respective teams or groups.

44:46 (Afri)
  And then just looking at.

44:48 (Afri)
  The calendar.

44:52 (Afri)
  Let'S meet again on July.

44:54 (Afri)
  13, then, which would be in four weeks.
  Unless you think we need to meet earlier.

45:05 (Jim McDonald)
  That feels like quite if we're aiming.

45:07 (Jim McDonald)
  For Genesis in September, that seems quite late.
  Maybe a couple of weeks time or something like that.

45:15 (Jim McDonald)
  I understand you're on holiday.

45:17 (Jim McDonald)
  After you but someone else can run the show.

45:22 (Afri)
  I don't mind if anyone else wants to.

45:24 (Jim McDonald)
  I think Barn was just volunteered.

45:26 (Jim McDonald)
  Right.
  There we go.

45:36 (Jim McDonald)
  Parry.

45:36 (Jim McDonald)
  Thank you.

45:38 (Barnabas Busa)
  Proto brought up a good question about whether Altus should also be running nodes.

45:44 (Barnabas Busa)
  And I think that's a very good point.

45:46 (Barnabas Busa)
  So we should probably try to push.

45:49 (Jim McDonald)
  Some Alto clients to also bring some notes to the table.

45:54 (Rémy Roy)
  So I think that's fine.
  Again, we need to remember that we're.

45:58 (Jim McDonald)
  Talking Genesis here, so lots of people can join after Genesis.

46:02 (Rémy Roy)
  So we shouldn't be fretting that we.

46:04 (Jim McDonald)
  Don'T have enough layer twos or DVTs or anything like that if they're not ready yet or it'll take them a.

46:09 (Jim McDonald)
  While to get ready.

46:10 (Jim McDonald)
  I don't see that being an issue.
  But I also don't think that we should delay Genesis.
  We should make sure the ETH is.

46:15 (Jim McDonald)
  Available for them to run validators when they're ready.

46:34 (Jim McDonald)
  Okay.

46:41 (Afri)
  I was distracted for 1 minute.
  Did we agree on a date for.

46:44 (Afri)
  An excellent two weeks?

46:51 (parithosh)
  Yeah, let's do two weeks.

47:00 (Jim McDonald)
  Okay.

47:01 (Barnabas Busa)
  By the 29th, we would want to have the hard commitment from the client.

47:05 (Barnabas Busa)
  Teams and possibly some other two.

47:08 (Jim McDonald)
  Right.

47:10 (Jim McDonald)
  What would be really good to have then, as well, which I have to say, because it's not my work, but it would be great if the client teams can give feedback on the impact.

47:19 (Jim McDonald)
  Of having a Genesis that's got, say.

47:22 (Jim McDonald)
  1.4 million validators in or something like that.

47:25 (Jim McDonald)
  Because obviously, if that's not feasible or.

47:28 (Jim McDonald)
  It just bloats the binaries too much.

47:29 (Jim McDonald)
  Or whatever else, that's something we might.

47:31 (Jim McDonald)
  Need to have to worry about.
  Now, we can obviously run external Genesis files for most of the clients, certainly.

47:38 (Jim McDonald)
  But that's then a pain and a.

47:40 (Jim McDonald)
  Bit less standard and all the rest.

47:41 (Jim McDonald)
  So I certainly think hearing from the.

47:43 (Jim McDonald)
  Client teams the impact of having a.

47:47 (Jim McDonald)
  Testnet Genesis with over a million validators.

47:49 (Jim McDonald)
  In would certainly be something that would be good to know.

47:54 (parithosh)
  Yeah, we can bring it up at AC DC immediately after this and already try and get answer there.

48:01 (Jim McDonald)
  Fantastic.

48:05 (Jim McDonald)
  Okay.

48:05 (Afri)
  Sounds good.
  I will publish our conclusions on GitHub.

48:09 (Afri)
  And supplement it with relevant links, and.

48:14 (Afri)
  Then we continue discussion on CD and.

48:16 (Afri)
  On the next call in two weeks.

48:23 (Jim McDonald)
  Thank you.

48:25 (Afri)
  Thank you.

48:27 (Afri)
  Have a nice day.
   Thank you.

48:28 (Barnabas Busa)
  For real.