All Core Devs - Consensus #36

Transcript

[Music] [Music]

[Music] [Music]

[Music] okay great the stream is likely transferred over if you're on YouTube let me know yes there's an agenda issue 135 generals similar structure and we're gonna have a little bit more in-depth conversation

about multi-client test nets after he's here who's been dating around I think you got this and we'll get to that a little bit okay so we have a release in the queue I anticipate that it will go out today this has some continued networking updates that we've been working through the past about 10 days it also has fixed virtue critical bugs one in testing and one in the rewards calculations a rewards was set to equal instead of plus equal so which also highlights some both of them highlighted

that it's likely valuable to split the attestation Delta's function into some sub functions for more granular testing because it's very difficult to capture much other than whether the values are going out there down but if you mess up a calculation in the middle with respect to testing it's easy to throw a little function and not really really cool so that's coming this release I'm because of the critical bugs and the previous one we're gonna go ahead and say use this as your target that's very especially respects the state transition very minor I hope y'all are having a

good time working through p0 11 I'm sure you let us know in a minute other updates on testing and release yep okay so we've been working on the Java integration which is pretty much done we got the shuffling up and running we just need a way to disable signature verification on take you I'm sure they'll come very very soon we're revamping the architecture completely we're actually considering rewriting the differential for any rust so we want to have the ability to swap between different fuzzing engines so we

currently only loosely follow but we want to be able to have a CLI that you know allows us to swap to AFL Conklin's Lane and potentially others so they all use different mutation angles and it'll be good to experiment with that with all of them we're building up our coverage recording to better measure the efficiency of the structural fuzzing which is coming together we're prototyping new closers using a lighthouse first and we're hoping to have a lot more progress to show these two folks in two weeks and they're excited to announce that we were onboarding a new person to the project Patrick wasn't security fuzzing expert started working with us we'll have a

look into e2 and all the different implementations writing started online on Monday we also found a minor type error in Isis bed which is not exploitable or at least not directly exploitable but we'll make sure to report it and that's it great thanks buddy proto anything on your end sorry anything on your inverter so stupid study progress on network testing a swimmer is if all thing like introduce more features we are starting to get a good interface in the Python repair room

so in the interface and use rumor for our sink last as a proof of concept we have synced the first 10,000 blocks of light as does nuts if to inspect specification including and so mr. only studying synchros it's pretty swim great thanks Berta other Abids okay let's move on to client updates let's start with TECA so we have optimized the American banking

news so that the default transition is now like four times as fast is what we have before we're now sitting at about 300 milliseconds vertical transitions work $33,000 with the main tip pumping we added a mode where prune finite state as a quick fix to keep stars from growing like crazy while we test thinking and we started but not completing work to replace mob TV with rock CD racks Libby is out of the box a bit faster than mob baby so we're expecting a speed-up and describes that our discreet file implementation seems to be working well and like testified and we're working on putting

out a separate bail very point the rest ideas are continuing to and we're hearing that like Senator standardization of API to be really useful so it came to that and we're hoping to find something with time and effort to like permit that and lastly we're working on a standalone signing service which support them live in peace and the hardware security module that's about it nice are you all a below chol test night right now with just technical or just a not a public Tessa I go reach that milestone and yeah we should be we

should be like

Cem Ozer: we should be able to do that

yet cool just kiss thank you prison yeah hey guys Terence from Chris Matthew labs so

Terence Tsao: we're basically just working on version 11 and with the biggest feature which is the dynamic at the station subnet that feature that it's complete right now so we just need extensive testing and then iron out a few bars here and there so so and I'm talking about where we have a slasher Rodney in production catching slash turbo offense and we cut our first it's a novella dated double booked last Friday and verify the external date / /

so that was pretty exciting we added our continues first casting as part of our CI pipeline we have seven first Argos right now which are tested with coverage based random inputs we also made a lot of radical improvements on on the advanced TB computation so we now we have a rapid service on top of our state evey to maintain a both of hot and cold an accessible state and we will star extensive testing on that later this week we did really good work on separate out initial sync clock that iam from to our processing we factor initial same service and then introduced a controller

model of syncing so meaning the plots are processed as they are being fetched and that's it from me

Danny Ryan: great thanks Terrence and Congrats on catching the slashing

introduction Libous the heart from Nimes team

Zahary Karadjov: we continued our work on the discovery we fight implementation and we're currently interrupting well with the go implementation and the lighthouse test net where I go to discover 130 nodes there we completely dropped the goal demon from our development branches now

and we've launched finalizing test net based on Lippe to be but the stabilization efforts are still underway then we are turning down some resource leaks so we are seeing in this test net we've synched with the latest spec dot 11 but we are still have some pending words on the networking changes like snappy or similar items with reduced our the json-rpc library we got one from our theorem one client an integrated team in the beacon oh so it's now supports some

of the RPC endpoints that are supposed supported by the lighthouse for example we are currently looking and trying to fix the scripts in the multi net repo so we can do local testing with all the plans hopefully maybe some of the other teams will be interested in this as well and we can bring these scripts up today we've done some work on syncing and currently sinking is working in the p2p a snap but there will be more ongoing work in making the algorithm smarter and

both bouncing to the requesting more appropriate way and finally we've been upgrading our compiler tuning one point point two which will be out in maybe the next few days

let's start ok so we

Cayman Nava: just finishing up to c5 implementation you need to integrate it and do a few more tests but it's good we've been spiffing up our two node local test net just fixing some

various things we noticed we had a bug in our gossip validation and we've been just fixing up our API for that and we've been working now on on spec updates and we got a new contributor recently who is helping us out with our like a web apps he's gonna be updating simple serialized calm and he recently is working on an eighth - validator key generator web app and that's nearing completion

Danny Ryan: lighthouse hello so this one I wiII be focusing playing

expect version 11 so the majority of the work has been around during the attestation aggregation scheme working mainly from a networking side stays Orion has been working on that a lot along with Hawaiian and diva so Adrienne those team have those people have put some new search routines in this V fire to assist with peer discovery on the sped up this v5 queries by adding timeouts and been thinking a lot about handling Forks and the networking stack michael has been working on implementing the state transition side of things he found that that add a sign bug which was pretty cool so good on him we've been do

feeling with the API a fair bit trying to between the validator client and the beacon node as we implement we figure out better ways we can do things we've been working with a free who has been playing with some of our tests at deployment tools just kind of the first time that we've had anyone use them that except for the altars so that's been really helpful we've been submitting lots of really good issues we did some more optimization stuff around batch verification of blocks yeah that's about it from us great thanks Paul Trinity

everyone nothing too crazy other than kind of the stuff for one else has been working on we've been working to pull out a separate other client so that work is pretty much on the completion which is exciting to see otherwise just work on node stability and the b11 spec updates along with some infrastructure work for single client

test nuts

Danny Ryan: great thanks Alex another mind you have anyone today don't cool something else that I meant to announce earlier is that

with this next spec update we're going to be relaunching a phase zero pre-launch bug bounty program so get some details on that out soon okay I got

every one right great any research updates yeah maybe I should give a since I haven't yeah but so here hi everyone yeah so we account on verification we

have been working on generalizing the Casper model to to support dynamic validate turns and and we have done that so we have what we have right now is is a model of Casper as given in the gasper of paper so so the model has books of of accountable safety and possible likeness we use and for assuming K finalization and dynamic validate cells and what we are working right now on is on showing

the bound given by the theorem and that big this is also in addition to the the second model in K so they the K model is an abstraction of the beacon chain transition function state transition function and that K model does actually both fixed very data sets and anna-marie datasets and it also includes partial refinement proofs showing that essentially linking the two model to the two models together showing that the K

model satisfies the assumptions that are made in the in the top level model via the croc model in addition to some properties relating to the state transition function there are still some some other properties that I think still need to be proved but we are working on those as well so the the point of this effort is is to show that the the we can change state transition function satisfies accountable safety and possible lightness that level and then the ultimate goal is to be able to eventually show that the implementation

itself satisfy satisfies these properties Thanks so that's yeah really exciting work seems like close yep cool are the research updates per take Eric we just had kind of maybe a realignment on some some research who are going to we're helping with these is that never helping quilt with the ease research and that's

a II framework and I think the general direction now is didn't lose away from these as a design for at the earn - no no it's going yet but um so I think Johnny ray is going to move into doing some sort of network monitor um and this would be helpful for say zero being and when I monitor the activity of the network and some better understanding what's going on there and mikail just published eight and some research about and East one to bridge finality gadgets and I think that are

his conclusions are coming closer to the conclusion that the I think the foundation is already at that is a bridge or a finality gadget is length to finality makes it kind of like an in feasibility and so I think we're gonna shift into working on them each wanting to merge more likely go thanks Joe any other research updates Sofra and I've talked about this and the

other calls before but regarding a polynomial commitments for stage there's been an obviously that each research purse that I published a couple of weeks ago and then since then Dan cratis public history man Dallas Terrace comp with Miss diem and the thing that I presented yesterday of this UK study quad was basically some improvements to the version that I came up with that remove the need for a sorting argument that so the number of constraints goes way down and the entire scheme becomes quite a bit more viable cool rapid progress thank you other research

updates great cool the next item is well tonight multi-client test net discussion largely there's two efforts the first of which is getting a multiplying test set up the which we might see a few give birth and die rebirth and then also to do sort of I sense if you could focus largely today

would be on that first effort after II do you want to talk about what you've been thinking we've been looking into your thoughts here

Afri Schoedon: yes thank you for having this is my first East Room 2.0 card so to be here I previously were no me yet parrot is not interested in the serum so I have plenty of time now and historically I've been following assumed to point out since early implementations officer hi Bert Casper proof of motive of stake and I believe Trinity harmony and parity and we've come a long way and now that assume 2.0 spec is more and more

stabilizing and I seasoned needs that there needs to be some correlation I stepped up to volunteer in cases wanted to help out this particular chestnuts and yeah Here I am now and I want to learn more about how we can we our security project can assist so client teams we have resources available I have plenty of time I can have coordinating things I used the last two weeks to investigate clients capabilities tooling around creating chestnuts what does it involve to create a test net and so on and yeah so yes polar I

said I mainly work with lighthouse but I also want to learn how all the other clients are capable of interoperability I know there has been an interoperability or a multi-client test net at Ford I think in a couple of months ago but we are now in a phase where we should work towards a more persistent version of multi-client test nets and yeah so I want to learn more about which version of the spec is implemented which client what's the start of networking consensus validating stripping what's the status

of tooling around creating test nets in for each of the clients for ads I'm not aware of and yeah this is all the things I want to learn more about I'm I'm q9f on this core and you can spin this out and to having separate cause on this we can use separate discs or shares of cases but yeah any any comments on that I'm happy to learn more about what can be done how we can contribute

yeah

Danny Ryan: cool yes thinking you and I can either go back and forth on some of the core requirements and things that we're gonna

need to standardize and then share that more broadly we've done surveys in the past to kind of get a handle on where everyone is at especially with respect to the core items so we might do some sort of limited survey again as for organization beyond this call that's likely warranted I think we'll probably at least a couple weeks out because I think everyone's a version that would like to target is pretty much the hadn't expected a and people are working to get there as discussed on the networking call yesterday some of the complexities

people are working through some of the complexities in this aggregation scheme I think prior to very recently we're kind of cutting corners there but we expect to be fully up to the spec very soon

Afri Schoedon: yeah I believe we need to also have a joint at some point we might want to have some strategy to freeze certain but I don't know what you have done previously and oh yeah I'm happy to know about that

Danny Ryan: yeah long ago prior to DEFCON before we should start to make some deeper breaking changes based on the conversations there we did have a spec

freeze the I'd like to with this spec update freeze the state transition logic unfortunately also fortunately we at least in the past two weeks have seen a lot of conversation around networking primarily because some people are working through it the first time and also hitting some the pain points for the first time so we've seen more iteration on that component of the spec and some of these iterations have been breaking so it'd be hard to unless we were versioning protocols within the same protocol would be hard to freeze that today I think likely in the next

two weeks we're still gonna see some some bumps on that so state transition freeze and targeting soon a networking freeze I don't have a good intuition on how much how many more changes we might see there because a lot of us just been led by the kind of networking experts on each team finding some of the issues but I know we're getting into a better place there

one other

Carl Beekhuizen: one other point which we may

end up changing and which I would like to see it possible change before we start doing

test answers the new via this fix and hash 256 did that sexy draft get released yes we have my version v6 and a hash to code to so there's a both your yeast I'm opening the second PR for again spicy see as we speak so from there we should be able to move pretty quickly the changes are minor for clients so I think we may see a reason why we shouldn't do that before not the Plantagenets minor we're talking the interfaces are all the same for the client but the the underlying logic yeah I I would almost recommend reading the

Python instead of the that nothing's changed basically in hashed opposed sorry indeed via their specs v2 basically the old ones expired so the update was just the version bump and the past occurred things have changed the hash the field mechanism has changed from HK DF - a sort of custom key extension also accustomed extension mechanism so that's yeah that should be doable and it was I think a bit ok so I think

Danny Ryan: what we'll do is this version releasing today free state transition

have it be the tangible target maybe a release of version 12 when we have this breaking signature scheme come in knowing that the rest of the items are stable and as clients get up to that up to speed you can shift the effort towards that

that seemed reasonable well yeah that's what I mind today

Danny Ryan: good that they got the v6 out and those contributors are saying stability or there it did anything else nope in the pipeline

nothing new expected from

the pipeline I would still like to have seen but the only shirt I'd like to have seen is kiss victims in the bitterest effects on them but yeah otherwise yeah that should be stable

Danny Ryan: we talked a little bit about this long there's a blog post and some of the chats the intention here is to not allow the ITF standard to be the blocker on living the main net and so we will try to implement their latest whatever the latest draft is and ultimately probably try to shake hands with a couple of other blockchain

projects that are also trying to move quickly and conform upon one of the traps if they have not fully stabilized but right so I think a lot of this conversation is probably going to be had on some different Doc's will be sharing significant things that will try to standardize on and increasingly so spending out test Nets testing sync across clients and things like that is there anything else after that you want to talk about today or is any questions people have for just about this process

or thoughts or comments before

anyone I mean yeah

Afri Schoedon: I have a lot of questions but I will be going from Senate Shannon from plant client and folks team so please be Rosemere if I have a lot of questions I I just need to learn more about where each client is and what the tooling status isn't how our chestnuts can be spun out each of the clients and yeah I will take those offline mainly but if anyone has a comment now just go ahead

thanks al for you and you know I think it has been valuable to lighthouses

after he's been poking around and they count having a dynamic conversation about this tooling so help them out and that will likely help you out too after he is are there others from the Goya initiative greater interest in this are you flying solo right now

Afri Schoedon: I'm solo right now I mean the golly initiative is it's just three people and yeah but we have some funding and we have myself and maybe in future we can even have staff bountied if necessary we can see what was required or where some incentives are required

we're gonna move on anything else okay

we had a networking call yesterday then has some great notes there's a couple of PRS that are up and under review based off that and we expect to get these and with the release today other items networking items concerns comments anything before we move on and if somebody has a link to yep thanks Ben Ben's networking that's your networking going once going twice

okay expect discussion we've touched on probably the most pressing items which are that there are a couple of political bug fixes that are coming out just keep your eye out on that and I recommend just immediately bump into that version instead of finishing honestly because you're gonna finish out on zero these are 11 I could be I'll pass the test because test generations beyond that on the phase one stuff Xiao Wei has been doing a deeper view and there's also been a number of additional test vectors added so once that deep

review is done and we really will hopefully really say a version of the phase 1 test vectors and get when it's people prototyping some of this work is going to be that all for doing somebody xx 2 integration prototyping work in the coming months so I'd say once the Xiao Wei fixes are integrated if you have some spare time to do some reading and are just kind of interested in getting caught up that would be a good time to check it out certainly I think that's most of the specs up on our end do other people have

any expect related items that they'd like to discuss I'm general local discussion I've been super pleased with the integration and the unification of direction roadmaps and research and expertise super super cool to see yeah I think I said last week anything else people want to talk about today nothing from me maybe we should do a virtual interrupt meaning at

least a few people from each team one week in the nearest future we spend a lot of our time hacking together on things rather than in our own little silos is that does that make sense is anybody interested in coordinating or in doing that I know our efforts when we all got together in person we're extremely fruitful obviously there's a degree of separation we're hanging out on the internet but again I think that might be fruitful

Alex Stokes: I think it makes sense

Danny Ryan: cool I'll chew on that and maybe propose some dates

cool anything else

cool these calls are easy when everyone's heads down working on their items so great thank you will likely do this call up so I'm not sure starting to put it at maybe two or three weeks but I'll keep you updated otherwise talk to you soon thank you everyone thank you thank you [Music] [Music]

[Music] [Music] [Music]

[Music] [Music] [Music]

[Music] [Music]

[Music] [Music]

[Music] [Music] [Music]

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