All Core Devs - Consensus #27

Transcript

[Music] [Music]

[Music] [Music]

[Music] sanskrit and to those on the YouTube chat box please let us know when you can hear us okay there is the agenda as usual we'll get

started with test net and release updates easier 9 o is out I think with us what we were discussing in the last call came out about a week and a half you know I know a number of you are working towards getting that cleaned up and ready to go and we have a few more PRS that are gonna be really so nine releases that I'm very very actually working on getting done and expect to release this weekend there are an additional cleanup of pulling the custody bits out of a two stations and a few for choice fixes that

are kind of post-audit fixes booyah a Japanese researcher has been doing some work and the past few months on III search has found a couple of attacks on fg+ ghost and a couple of pretty simple fixes so those are gonna be integrated again I'm working on getting these out as soon as possible to note that's really the delaying component on living towards larger distance I have a on our agenda today a discussion point for test sense again I expect there's some movement on the more localized test nets and smaller client test sets but on

larger Orca shared the ones we gotta get these I think we really need to get these releases out before we move there I want to make a clarification on the [Music] tests so we did have a number of these like ad hoc Interop test cases that we found debugging multi-client test nets on the wire these were 408 three and I think some of you had integrated them into your test set testing flows we have since then built out handwritten tests that hit these test cases so if you're passing Oh 8 4 and if you're passing oh 9 you are passing those tests so you don't need to

integrate those into your client anymore but we will likely run into those again and maybe do the same thing any other testing updates per duty of any updates on Muskoka itself up website this life problem is the V now needs worker notes or these clients to run on the sides and consume these inputs and I'm in the process of deploying the specification during these transitions and then next up i spec maybe n COI and now nci's working well and absolutely nice during interim so I'll

try and apply it as well ooh and that that end up correct me if I'm wrong that might serve the purpose of kind of cataloging some of these tests like broken cases right so the main goal of Muskoka is to we're in the same transition of all these different clients and make it except us XS boss possible for someone to try and win their test case all the different clients about like downloading the clients installing winning it themselves and then we publish the results and so

it can track what's happening and if there is any case where the results are different yeah so for example if I'm running some inner test cases against Nimbus and we have a consensus error I can post that to Muskoka and see how everyone else like either agrees or disagrees with our results yeah we're going to some consensus with Nimbus that that's always always the truth mommy but uh maybe you have to interrupt y'all clean some things up cool so want to any other testing updates I think this is what

okay okay thank you let's move on to client updates and since we're already talking have a Nimbus okay so what I said about consensus with moon this is actually not true because this week we are moving from zero point eight point three to zero so depending on the funding that you use the genesis state might change due to the change in the become state so Nimbus is not a consensus with Nimbus otherwise we updated abducts so we have no build instruction for building Nimbus on the raspberry pi for the public test

max sorry I've created new repo on the if two clients come on which was attorney for everyone I I put the link in the chat and guidelines on I mean everyone can other is own questions or change man just use common sense on how to share the metadata needed by your client on if another client wants to connect to it and regarding public test net so we are setting up infrastructures so that we have monitoring public monitoring through a graphing a dodgeball and over

the next weeks otherwise so like I said on the core part of the client we are dating to 0.9.0 and passing test one by one that also means that we implement both specs is repond 8.3 at this and report 9.0 at the same time and you shouldn't rely on States during the immigration because sometimes it would be zero point eight point three that you gets realized I was reported at 9.0 regarding testing over the past two weeks we struggled with time outs in OCI so we solve that

yesterday we have the caching issues of Rods DB and we also moved away from ab veil to other pipelines for Windows testing and we also have now testing on arm devices of all over comets and lastly we moved completely to tar bones we used caching for the LFS EF files in the past but now even if someone does a bit clone and FS something it will not use for sure the NFS quota of the foundation we finished also are running

out the bugs and grating to name one point O which was released last September and on the to be front we had some updates on observe and go sub implementation and now we are starting to we will start to integrate the natively p2p into a code base and replace the gold p2p demon on those the test net standard under the if if p.m. is there a place for specifying the version that the test site is targeting the spec version eyelids re replied because these

so the one who forgot about everything each death net existence of folder and it's a folder has a readme file within this video file you can describe any particular area stinks like which you are targeting whether you do the NAIC or other station aggregation and so on okay that's reasonable thanks okay how about let's start yeah so past few weeks we finished pulling out our state transition into a separate package this is gonna help just with people at a

hackathon I want you to use a state transition and being able to use independently of lodestar we are almost done with a more robust and official syncing we're in the process of moving from 80.8 inspect 0.9 pretty close to finishing on that we are refactoring our gossip sub because J s Lib p2p is going through a very large refactor from like an older style of JavaScript using callbacks to the newer style using async await it promises and so someone from

the lib p2p team is helping us with that migration and hope that that can be promoted soon and I believe that's it for now cool thanks and yeah if y'all didn't see there's a number of these two related bounties at Pete's Waterloo coming up and if you have any ideas for these I think we're gonna run them generally at the youth global hackathons if you have any ideas for little experiments or tools or things you think that would be good targets please let me know and I can get them on the list great how about parody so sir

first for us for party we finished Rosa we fix all the the Interop usual we had since last time so now it's ours it should work and we also finished upgrading our beacon implementations who to open night hungry just passed all the reference test today so that's for us great thanks way Trinity everyone let's see so we have a lot of exciting stuff and PRS that's really just waiting to be merged that's the state-transition updates 2:09 and the network updates from zero for we

have a lot of good work on node stability and pile of p2p stability so those things just work better we had a really crazy performance improvement in pajas has said so moving basically to have backing persistent data structures so we don't have to like do well crazy copy and we were doing it had something like several orders of magnitude increase the performance so that was pretty awesome and then some work on separating out our about our client has a distinct binary and very much looking forward to all these product tests that's happening as soon as possible great thinking thanks Alex lighthouse hey everyone

same as everybody else's that we're working towards doing public test nets so in that regard we've been focusing on our s1 connectivity which is kind of helped us clear up some of the issues with one voting speck we're making changes to our CLI to like make it more organ amah so that it's easy to use when enjoying some of these test notes we've been ramping up a bit of our testing as well and most of the aspects of white house going for actually getting public use we also don't we've implemented encryption for BL s keys so we have like a key store forever DK's we're passing the test SPECT is for 0.9.0 but we still need to kind of

finish the update which will probably be doing next week early next week we've also been implementing slashing protection we have actually a PR fashion protection for the validator client and we're looking to use SQLite for the database inside that I think that's mainly the main stuff from us thanks.thanks age and slashing protection there is ensuring that the validator doesn't get slashed not policing the other validators right that's correct yeah prismatic oh hey guys and it's pretty much the same all day plus everyone else we face if you are PC related parts after users have been

before the team name so which is nice that today we have users querying our testing information and the our single client testing is currently done right now since were lining rushing through 0.9 to our database so we finished which means during on a few days ago and it will cause people to spread test for both stations to Sherry and si si so those look good so far and we're planning to relaunch our test name by the end of the week and in the process we're also working on aggregation but that's not really passive party we can run in the newer aggregations batch as a teacher and and was always you know splashing on both protection and then and then the

policing forefront and yeah I think that's many it great thanks Terrence Artemis you know so for something a little different we've got some teen changes removing Ultimates internally from incubation to the product side of Pegasus though it'll still be some familiar faces but there are one or two new ones I'd like to introduce Meredith things on the call she did a lot of the heavy lifting on our theory one client and we're very glad to have her joining the Artemis team as far as development where it goes similar to everyone else this sprint is about updating 20.9 we're

completing discovery v5 integration we're working on the rest management API and working on syncing and the goal is to be able to hook up as soon as possible to some of these Testaments that you're the building that's it for me thanks for welcome Meredith and do we have anyone from harmony again I know there's a unification in process but harmonies kind of driving forward on some more research you related components so if anyone from farming likes give an update so what about the Harmony we have updated we're officially

a gated participation to complete vultures this we where interruption with yet for eg scour a g5 we are continuing Piketty the simulation and so we are continuing to okay thanks anthem did I hit everyone okay jinda next we have research updates the tale excited

be about a half-hour late but we can we'll do some research updates and if you joins us give us update anyone get us started I don't know I'm gonna yesterday we start had our first meeting with light client friendly folks the light client task force soon directly I guess you know nothing new presented but we had a great Q&A with the Schultz all foldy on the like line and sanitization lies client-server and sanitization framework that he's come up with we have a recording online it's an hour long so I'm gonna see if we can get

a grant or not a grant a little uh bounty together to get a transcript but it's great to just get get this this uh knowledge out there to everybody so if you have time extra hour you can listen to it well thanks Kevin Justin yeah so we've we've made some progress on the Els civilization there's a new hash to curve draft which is out and this is the one you know that I'm expecting we will go in production wave boy theorem there's also kind of updated test Suites and he Rumi so they make the maintain of her

room either the really fast GLS implementation has accepted to do some work on the grant so he'll be implementing this new hash function and steps and that's all my updates on kind of phase zero but if we look into the future on phase two I guess there's there's been you know very good progress on the you know the vdf project I'm like one of the big milestones is this RSA MPC and so it looks like you know the MPC is hitting all the

performance targets so we're looking to [Music] wrap up well handle over the code from the hero that is implementing the MPC and then hand it over to auditors maybe early early next year and then hopefully do the MPC like something maybe Mittman next year but it all seems to be working moving very well and just generally the the the space of no groups of unknown order the the cryptography seems to be blossoming because we have these these new constructions for snarks for

accumulators and there's also new groups of unknown order that are rumored so beyond the RSA groups and the class groups there's kind of a rumor the yet another one which has some very nice properties so yeah this is exciting stuff that's happening that will be relevant in Phase two and also at the application layer great thanks Justin any other research updates I have a question for Justin hurry we have the rest integration but I guess the base

level will be in C++ with bindings as well right yeah so I mean I guess if you have a C project or C++ project integration should be easy is that the question yes it's just to know if we will implement it in rusts or in CEO stress with and we will help with bindings from sin to worst for example aha well I guess this there is a telegram group which I can add you there I think this is kind of still being discussed with the Sigma pine guys who are helping

with that integration more than happy to include you in the take member yes thank you any other research items before we move on so maybe one thing like we've made some efforts over the last couple of weeks to unify like also in the similar way nice question what Justin just mentioned on the different crypto stuff and to like unify our bounties on that so we now have a central entry point challenges that we framed at all and yeah I mean we're like expanding our

programs there because we've seen like some very good work come out of it and so and it seems like those bounties are a good way of like getting our cryptography and we're doing some very new cryptography here at the moment yeah better validated so yeah if you have any people who are interested in that you want to share those with like challengers 33m that are our good entry point to to see all those different bounteous that are yeah there to improve the state of the etherium infrastructure yeah I just drop a link in the chat thanks I'm good I forgot we forgot one update I post in

the chat just now but with test nuts coming up where I've been working on like a meta mask plug-in if you don't know it's like a new v2 kind of thing I'm a damask but anyway it didn't look like auto deposits and stuff like that like built into it so that hopefully for the getting users on the test fence and stuff it'll be a bit easier and less painful nice on the auto generated deposits are you gonna generate keys there and then allow them to be exported or what yeah that's the idea the meta mess lets you like derive a key from like the root key right and then you can use that as like salt or

something to then generate or entropy to like generate a new pls key there nice yeah on that note does anyone have confirmation because you're kind of debating it how much overlap is there on the current eath one key to the it's curve sorry to the e to the BLS curve like the I think it's like 50 percent right it's not it's not hundred percent ok cool so definitely not gonna use it I keep that entropy yeah have you seen car release a number of here yes yeah I peas which are basically in especially

living in my memory foes but nausea keys 23:33 23 34 and 23 35 which are the LSC derivation key parts and basically is if you if you want functions or mapping from entropy then you may as well use the HK DF Matar that should come to subdue you nicely and then be compatible with everything but yeah

I've had a lot of people in contact with me because some teams already implementing that full ounce of bugs in my implementation and unplugs all that go all the way back to test vectors for the one key source which is being little uninteresting but yeah good good thank you you're just you're bad at zoom great it's not my restaurant see any other research items before the blog great Network update I was

antoine anton from harmonies and working on a simulator for gospel sub with the DP pub stub but he gave us an update about that earlier is there anything else you wanted to let us know about that on Anton I'm just optimizing the code and for now now it looks like hundreds of thousands now candy can be easily simulated not not too fast maybe

it's possible the work is still in progress cool thank you I think Mike was planning on joining us from protocol labs but I don't see him right no one's from protocol abscess here okay any updates from white block nope not any nothing too big this week that I know of we're just continuing with our adjusting our methodology for the Lib PGP testing based on some of the Cebu cotton great thank you any other network related things I discussed

okay moving on the next item is general spec discussion I put that there we have the O nine stuff and the other nine one coming any questions that came up while you're working through that it's largely subtractive be straightforward okay good next up tests not discussion my general read on the past two weeks is that

people are working on oh nine people are also working on getting some teams working on getting some continued single client test set stuff out there and there's a lot of general intention to experiment with some multi client behavior on different teams single client test sets in the coming foxman two weeks like I said at the beginning the call a more orchestrated test net it's not something that I want to do until we have what looks like an e respect freeze with these oh nine updates coming out I also posted an issue on e 2 p.m.

yesterday that is just a list of tooling and things that we want for test nuts and also the Associated people that are have either mentioned that they're working on it or or not I just share that there please add some items that list and we can figure out how to we'll prioritize things and also bring in some external resources if we did decide cool so that's the at least from my perspective the Leyland on test nuts is pretty much we need another two weeks to do single client test settings and get these

dates out as there any other things open discussion on test nets comments thoughts on what's going on because it makes sense to put a test net configuration like with all the different parameters set in the in the specs or is it too premature could that go like with like Oh 91 or something right we had discussed at least isolating the I think some of the signature domains or the versioning so

that might be desirable I think at least from in terms of shark counts and things like that the intention is generally to do minimal and then to step up to the main net do you have any did you have any other anything else in mind for altering the configurations for test nuts I think I saw in the chat someone was talking about changing the eath one follow distance so the may want basically use the e2 stats for the minimum config and

then the other ones we change other are the EEPROM power distance and maybe each attaching powers and those are the two that which changed right ejection balance is definitely a good one what is the follow distance on minimal right now or is it not is not specified to change it's not so basic by but we change you from 1024 to 16 which is about 15 minutes when they give us a carrot a snail all right okay I'll look into be making a modified config called small test now that can make some of these adjustments

hey Danny do you want to go over the your new key arc with the deposit sighs yeah let's I saw that Justin had made at least thrown up some counter-arguments for discussion so long ago the number of validators to start the chain was specified as think about half a million this was probably about a year ago and then Alexei doing some just kind of getting away land and doing some reading

when he too was concerned about a gate keeper attack in which a someone with around 1/2 million he quickly deposits triggers the chain to start and then prevents new deposits from being processed this was this attack was in the context of there only being an amount of depositors to trigger the chain start rather than also a time so given that there's there's a Genesis time and given that so that it gives other people other than attacker a chance to make deposits if that was the

case that mitigates a lot of the attack potential and also if a gatekeeper attack happened it would be incredibly obvious and an opportunity to utilize some social coordination to remove an attacker from the chain which is kind of one of those near last resort techniques in group of stake to recover so I propose an MPR one four six seven to reduce the and we up in responses the the NICU attack we up to the min deposits to two

million validators are no two million sorry I was speaking validator to essentially reduce the chances that somebody some whale could execute such an attack more recently we've discussed that maybe it makes sense to reduce the amount of minimum beef to start the chain primarily in light of this like potentially long an unknown lockup period it allows for if there's if there's not enough alligators that are encouraged to to validate allows for a smaller number of alligators to join at a at a higher game in those kind of early days there are some counter

arguments posted by Justin I think primarily around the fact that Edgware had very high walk-ups very high lockup amounts in the even with 12 months of unknown you know 12 months of lockup and things like that anyway I'm generally in favor just because even if we hit this target too low and a lot of people show up to validate that's fine but this just gives us some more wiggle room and if validation is not super encouraged because of that lack of time it gives people a chance to earn a higher return for being an early adopter [Music]

anyway you can go check out the PR Justin you can present your counter arguments if you'd like or would you just take the discussion there but Craig is there anything else that you want to discuss related to this no I'm just curious to hear mark cuz I'll lock up more girly and we need to should it not go through really well yeah cool Justin yeah I don't care too much either way it was mostly just for this casual assistant great

so I guess my my grief on the lay of land for test nets I hit on the nose and there's no further discussion oh and it turns out Ben Edgington is those also a girly well if they drink Adrienne's away Wow okay cool move on to the next item just general open discussion any anything any losses ooh dress okay short and sweet meeting I know

we're all pretty heads down trying to get stuff done so I'll let you get back to it I believe we're planning on meeting in two weeks I don't think that's us Thanksgiving so I think I think we're gonna miss that which is good cool thanks everyone talk to you soon thank you right right [Music] [Music]

[Music] [Music]

[Music] [Music] you

[Music] [Music]

[Music] [Music]

[Music] [Music]

[Music] [Music] [Music]

you