All Core Devs - Consensus #28

Transcript

[Music] [Music] [Music] [Music]

[Music] you

[Music] [Music] [Music]

[Music] [Music]

[Music] okay welcome everyone we'll get started with today's meeting the agenda is in the chat those of you listening in on YouTube please let us know when UCS is live we'll get started

testing and release updates as for testing I think generally staple although I think Maddie will give us an update in a moment on some work that he's been leading as for release I do have a pending release on github with some minor clarifications and changes I think specifically clarify some things in the last update about subnets and stuff I'm looking to get that out as soon as possible I don't think that changes might change one test version

but I'll let y'all know that's coming other than that I think good Meddy has an update for us today Thanks hi everyone so as some of you may have seen Danny's update suma Prime has been awarded a grant from the you theorem foundation we do welcome because professor I'm very excited to keep contributing to the issue security ecosystem beyond I guess our work on the lighthouse so we'll be ramping up our efforts on first starting from next week getting more time and resources so far since the last update we've managed to

clean up the buzzing targets there are no longer relevant so we've just basically got rid of the function of the fuzzing targets that are no longer endure in spec we've also added a couple more block state transitions to our fuzzers so in fact good news all the block state transitions are now included in which is cool and we've also added to Hawk state transition functions that are running locally on my machine at the moment so looking at process rewards and penalties and process slashings we have been working on integrating other implementations as well so we've reached out to a couple of you folks so as a

reminder currently beacon first works fully supports high spec guarantee on lighthouse we've reached out to prism the build process not you know very straightforward but we got there at the end happy to say that prism is now integrated but unfortunately we're facing issue issues since we're having now to go implementations so we're facing a bunch of namespace classes so we've reached dr. proto who's providing this with a bunch of different I guess avenues and up shows that we can explore but I'm fairly confident that will we get there very very soon we reached out to name this as well so we are I guess trying to see how we could reduce the

static sea library then we could simply link to our differential fuzzer and we started integrating Trinity as well which should be very very straightforward so we started working on prison thinking that since we had the rnc up and running with the very I guess trivial to add another go implementation turned out to be quite the opposite because we already have a go pigmentation it's it's actually not not as simple as as expected next steps will be to finalize all the epochs their transitions so that should be done next week we'll also bump all the current fuzzers to the latest version of the spec not just sure exactly what the

progress is from the implementers on being compatible with zero point nine point one but I guess we're an update from everyone today and will also continue integrating eternity so hoping to have prism Trinity and Nimbus up and running on beacon flowers by the next update thank you thanks Mary really appreciate I'll take it out on

Danny Ryan: any other testing updates great I know there is a desire and even a need for for choices that's been on the back

burner for us for a while and you need to prioritize it I'll try to get some at least notes for review up on that within the next week I guess the good thing is is that some of these corner cases and things that we've integrated into the spec recently are those they are corner cases and not expect it to be seen especially on test sets and even primarily not even seen on

minutes ok so let's move on to client updates we can start with cosmetic yeah hey guys

historians from prismatic so we relaunch our test network version 0.9 last week so there has been lots of public interest there are currently 55 years right now and when we count them ours so it's deafening great to see that users training now also with that it also comes with other questions support and great feedback on the community so we have been improving upon those feedback and then also improving upon or testability and then we have lots of bad faces on the are on the on the RBC back-end because there's two big names blog explorers using them right now so

we have been facing the RPC backing as well now on the side we're working on aggregator of the world which was previously for us every validator aggregate and then broadcast the attestation so we're switching to these aggregators so that shouldn't project us from the Fisher's your opponents we're making SSD performance improvements right now because with one of our bigger bottleneck hearing I'm seeking also where we REM him they eat one data implementation I'll trade a precedent had a few questions on that so do you see that happening and then we also resume finish our foodie end-to-end tests from depositing the deposit

contract and then from the audience of submitting a deposit concession and the finalizing epoch so so drive him on that and yeah that's that's about it great Thank You Terence so the that current test that does not use aggregation as I think you know segregation

Terence Tsao: but right now I'm every validator is an aggregator

Danny Ryan: okay great have you put the the configuration of that test net into that shared repo with

any notes about about it

Terence Tsao: not yet it should be the same as

minium and confit with the exception of shorter heat one denunciation and not good if and

Danny Ryan: when you get a chance dropping it in there just us getting in the habit of kind of documenting these nets so that we can begin to

yep all right so we have [Music]

where we have a test that running on running in the cloud the orchestration and monitoring we're just not quite ready to push it to the public yet because we want to finish our documentation and we really came to get the hot/cold database in there before we make it public just so we're not filling up people's machines we're also working on refactoring I validate a client would be quite a big refactor of it we're focusing on acing corporations and we're putting in some new slashing protection stuff and we're also doing some aggregation in the validator client so if you are running very large numbers like tens hundreds of thousands in one machine this should make it much easier

for your bacon node to handle it some might say they were just trying to beaten him in the next round or tweaking around the most validators but I wouldn't say that we're also working with Harumi implementation we're still gonna keep working on ours just the sake of diversity I think we might kind of put them behind compiled flags and kind of see how they go because own room is fast and extremely impressive but it is night assembly so it's a tiny bit scary but very very very much cool for the networking side we've started working on noise handshaking lead p2p we've added civil resistance to the DHT and we're focusing on a testing effort to test all

Network components which is not trivial but something we definitely have to do that's it from me

Danny Ryan: thank you Paul

how about lodestar I think I'm only one from the lowest er so I guess I'll deliver the update so the most important part we implemented all the carl's apes in javascript and we are currently testing it across the environments and we will release it soon we also made a huge progress on this cui v in javascript and our 0.9 spec is pretty much done they're just waiting

the next release of jelly p2p because it introduces daesik iterate iterators and stuff which is kind of needed for that spec version and that's pretty much working that I can remember I want to really expect this to be delivering the update oh good thank you Felix had a question for you Paul wait how did you add civil resistance to the DHT that's one for Adrienne hey yes I think if it's for Felix then on the district 5 there was a few suggestions but essentially we limit

the IP addresses on a - 24 subnet so in a particular bucket we I think we only allow 2 - training for subnets and in the entire like on the same - 24 subnet and then in the entire DHT we limit the number of expect so shows link we have now weekly test net setup in place so we

reboot that is net every Tuesday morning and it's stable for a week and we merge we have a separate developer and we merge everything on Monday so you can use that as a stable to unstable branch - compared to Nimbus regarding the spec we are now compatible with your pod 9.1 we're waiting for the pod 9.2 and we verified that we have the same Genesis as this is report 9.1 and so that would be deployed on the test net next Tuesday we do have complex sync issue that is

being debugged and we working on better logs because while going through all the logs is a quite stressing for set up and even for us regarding crypto and BLS I did some benchmark of every new BLS on the Raspberry Pi without openin CSL and GMP so it was twenty one millisecond for pairing otherwise on next to do is working with Sigma Prime on be confessed so probably will add some books to n CLI so that it's easy to call or to maybe produce capi for phasing on the

networking front we're working well always on this three five an active one and side note on f-14 so Nimbus if one we finished Istanbul at fork implementation and we started working on a Grove wrapper to switch part of the status client to Nimbus if one instead of gifts and that's all fun great thank you and those chestnuts that you are doing every week is that something that the public can join

Mamy Ratsimbazafy: yes that's the idea

thank you one more question in the O

nine one are y'all using the aggregation strategy specified

Mamy Ratsimbazafy: no we don't have the aggregation strategy yet one thing also is that while we pass the minimal tests we still have some issues with version

Danny Ryan: thank you

way is your parity year so software for party we are mostly doing some mob foundation work for beacon Chen and the beacon beacon from state transition implementation and mostly I

think the conference is try to revive the subject run time which was unfortunately removing the pastor who the suspect changing completely and also subject anything too quickly so I do catch every the schedule by removing that but I think right now is very important we revive that and the goal is to hopefully first make a breach and second hopefully make some of the components like this time we try to build it in the using the subgrade runtime will do which plates

will try to hopefully also produce something useful for efforts for the runtime modules and yeah I think that's class

Danny Ryan: great thank you

a Trinity everyone so mainly

Alex Stokes: working on spec updates and performance to various bits so that'd be like the version zero nine one updates which are mostly all together there's really just some refactoring we had to do on the fork choice that's blocking being fully up-to-date there although I do believe we're passing all of the

fixtures the the spec test that is some other things updates to pilot p2p stability we merged in our eighth one monitor some great performance work on passes said which had been a big bottleneck for us and then we've started work on attestation

aggregation Google things though its Artemis I were were split into two teams now we're working in one team were calling Advanced Research which is what we're kind of like combining it Armony we're going to be working

client not been working offline and we have another team working on Artemis this for being a kind of like production readiness and and they're kind of making changes and implementing sync right now okay Johan says he'll give the update for them cool Hey so on the Artemis side we've finished 0.9 oh and we are just finalized that finalizing the naive aggregation stuff and then we'll take a look at nine point one right after

certain point nine point one we've got all the request response methods done to at least basic functionality we're just finalizing a review and then we're also starting a round of in drop testing with other clients if there are clients that have donker and they just published that would be very handy for us to help on our and drop testing we're also starting to investigate blog sink approaches it's especially hers mathematics and we're also working on this v5 focused on the intro side of that and that's it for our update great thanks Johan and

Danny Ryan: Meredith I saw your question in the chat I'll get

to detail there but the it is a function of the size of the validators that I don't know if that is explicitly written down anywhere but we can

both specifically sorry this is the the weeks of heavy duty period how it's dynamic for many use cases of a big city period we would instead just rely upon some sort of upper bound rather than using even if it even or drops lower for the safety health and network it's just much simpler to err on the side of conservative on it but I will drop some more notes on that after this call awesome thanks harmony

Mikhail Kalinin: there is a little progress been

done on gossip sub simulation is a simulator now able to handle it 100,000 of nodes and disseminate a message within there in a same amount of time further steps are to add mesh topology and add metrics like bandwidth and so forth to this piece of software next we did a partial update of all this back to 0.9 and this is within our work on the fork choice tests we are about to

finish this fork choice test for our code base and it's good to know that you Denny are going to to have some updates yeah let's not single no choice yeah yeah sure we we probably can use our code to generate some may generate some basic vectors also there is a work yes you have to be finished with just by 5 it's now been testing with the gap and some like reach Interop stuff the dimitri is working on i guess it's

going to finish it this week also I've made the write-up about station condition detection it's compiled from the ideas that previously being posted by by proto and by Danny and some other stuff that I've been thinking about around and yeah it's like contains some also estimation of sizes so the indexes and could useful for those who are design in this part of of the client I'll post it like tomorrow I guess I need some to do some polishing and there there is no much

progress on our side these two weeks s we spent some cycles on merchants teams yeah the next two weeks next couple of

weeks gonna be more productive thanks everyone great thanks we go okay I think I hit everyone right that's thanks okay okay we move on to some quick research updates italic you're on I start us off but yes I can hear okay great let's see

I guess

Vitalik Buterin: protocol side phase one and we've done some edits to the phase one speck and Frodo has been doing some work on kind of plugging it into all of the all of the testing machinery I don't think there has been kind of too many changes mostly and a lot of kind of fairly small tweaks the one and the one thing that continues to kind of not be done not be done in fraud groups and for me that's basically just kind of waiting on that Israel the the phase one design and it is being accepted by more people

one thing that we did talk about is a blocks containing multiple chunks versus the ability the ability to and have created multiple blocks and once a lot and then like basically the idea there is to just collapse the two-dimensional array into one and like that's simpler in some ways but it's also more complex animal s8a efficient and and in other ways but so but I think it does still look I know she feel like like like there might be an opportunity to and if conceptually snowball I think quite a bit and otherwise that's the main thing that could be added on as one

broad groups and the other groups of incorrect group of custody bits which are not too difficult because it's something because it's only in only a single round game but there's still like I knew economic insight incentive nuances there to think about I've been getting the feeling from kind of conversations with other people in these two groups that I think people believe more and more that data availability of proofs actually they are something important and are not just a

kind of optional extra to throw on and to throw on at some point like people are basically actually afraid of the possibility that committees will break them that committees will break if great gift rating - maybe skin break can cause the entire chain to finalize something invalid and phase one has been designed to kind of from the from the start with phase two and or with the data availability proofs in mind so one of the reasons well the main reason why blocks have multiple chunks instead of one root is so that it'd be that way it makes it easier to and if combined all of those pieces together into a single big root that but in order to but then after that there's the question of what

specific data availability scheme do we use and the two realistic alternatives either something fraught with based or something dark face there was a version of something fraud proof base that I coded up like almost a year ago and it would need to be updated for things like generalize the busiest and so forth but it via reticle he exists and the Stark based one is and if protocol wise conceptually simpler except there's this one kind of very self-contained much more complex piece

which is free which is actually proving this and verifying the start but the main challenge there is we need to just make sure we have a decent stark start friendly hash function and four on the hash function side there is also one other new ones which is basically do we go with a binary field based hash or a prime field based hash and if we go with a binary field based hash then like there's no problems but on the other hands they start people have been mostly focusing on prime fields rather rather than buying a binary field so we risk kind of being left out of infrastructure

and on the prime field base side the challenge is basically that the 256 sides of a chunk it doesn't perfectly match up with any particular prime modulus the the one trick I came up with to try to alleviate that is that blocks would basically come with a tag that just says like what number you you should offset everything by and if you do that what's my rub with high probabilities you will be able to always find the tag that works and it will have probability the first tag you randomly choose will be it will be a tag that works so the idea there would be to just kind of rotate every week so that no no none of the weaves fall into like that tiny small range that's not not covered

by whatever's prime right slightly blow to little bitty six futures so that would needs to be you kind of put into SS said in some form so I think next steps would be to kind [Music] I mean one of them I know is that the Incred wants the protocol who wants to prototype this dark based stuff and kind of see how actually viable it is and I'm in Asia bytes per second sense and otherwise there's the option of just

like figuring out how to kind of swatting but whatever version of the proof videos sense what how doesn't want to be it incorrectly oh yeah those are the me any kind of protocol the protocol level things from my site so far and otherwise my like my own time personally I've been spending it on some hath like application layer questions

and in the last week as well

Danny Ryan: thank you for telling those stark friendly hash functions that were relying upon those don't those aren't those have not yet been vetted by the community area of

comparators I mean

Vitalik Buterin: this is basically like the entire reason why you like however i'm stark friendly hash function and our grant thus dark where and all of these other kind of things that have been in motion for over a year exist and they've provided some where I know candidates that we can use with medium confidence right now but in in general people like there's also desire to be conservative

there was a question from the chat any news on the ITF forum in which the BOS standardization was

formally proposed carl do you have any updates on that yesterday morning Chris Wood presented the hash to curve idea meeting it was mostly sort of just update I guess or for those people explaining the changes that have gone in since the last meeting which I believe was at three three so that was people won't be five that explained and then some questions around where to go from here and all of all looks very good no one raised any issues which was sort

of precondition on making this a blockchain standard so in terms of the the hash to curve it so the next things are happening is the sage proof of concept which is for the be master the most implementation that's being used to generate the test vectors which were removed for B fives of a during the test vectors back in making sure all the canonical implementations of hash to curve or good again which are the ones in go Russ let's see I've ever written which should all happen in the next two weeks or so but those aside there should be no surprises there so it's pretty

like we need to wait for anything and so that means that they're going to move this on to gloss call status on the an ITF standpoint and will be submitted to the CFO G for input on the cryptography but again nothing expected there that's more formality I think for my understanding how long is versatile that's so much longer so that's all waiting for this to be a proper a proper standard in in in my view that's six months off or something okay so it's not something we should be waiting for and as a blockchain standard these sort of agreement we beach from the he lost all

we had was a much shorter time scale so basically after this meeting if one raised anything which Cohen did so the with with the reservation that nothing's being said we haven't had proper discussions about this since the ITF meeting and you don't have any like formal written agreement surface is a standard I'm pretty happy to reasonably happy to declare those to be the easier less not going going forward and don't really expect anything too change there's an additional set more test vectors which is always a good thing so yeah I think for the most cases then it's fair to say that the peer list thing is not really holding us back

that's for coming to proper treatment that's us in the standard but only no more after our next internal meeting which is happening but even about two weeks time but for now I think we can basically think of it as a standard

Danny Ryan: great and there's a is there a PR up for review

and the specs we think there's a purview and I've been through it that looks that looks pretty good for the Trudeau's updating the hash the curve to be five it needs some of the definitions the specifications of the in the RFC

specifications I don't know whether we want to take those off and sort of explain them again in pseudocode or Python in be in the specs or not but what what is there is is is is good okay from what I can see so yeah exciting so I think progress there think those things already looking up any other updates on your end no I'm gonna squash all the BP commits on these those yeah keys and hopefully get those merged in this week great in the IPS are the Peter evasion in standards right

yeah yeah 23 33 34 and 35 grand just that's over if you want to know exactly what's happening and check that off we have a PR waiting to land with those key store implementations that's being great thanks awesome

Danny Ryan: cool thank you very much any other research

updates I have a few updates so zero I guess is the different foundation is kind of

Justin Drake: looking for someone to you

know manage the validators that we're gonna write so part of the plan is for part of the if the different foundation holes to be validating so we're looking for someone to you know we will develop skills with operational security skills to help set up high high uptime high resiliency high security kind of set up you know potentially multi-client multi-cloud essentially been using the threshold BLS signatures and provide all this infrastructure as as open source you know potentially being complimentary to you know institutional great

infrastructure that coinbase might provide so if you know anyone who might fit the bill please send me a message

like I mean there's an interesting team the community yes right so I mean we could I will use the DevOps team if they have had a strong bandwidth and if they're if it's appropriate we can also look outside I guess you know it's not mutually exclusive relevant for potentially phase one I I've been

spending quite a bit of my spare time on on zero knowledge proof and one of the cool things that Dan Boneh came up with a really neat secretly their election mechanism and he he wrote a paper and his paper got accepted so he will be published very soon and it it turns out that kind of the the circuit is and very simple so you could do it either large proof or you could do it as a foolproof kind of very similar to the date availability situation and of course if we can do it with his

knowledge proof is its cleaner and it turns out that in this specific instance it's you know potentially very clean to do it and very doable to do how much proof and then kind of relevant to face to you and kind of giving a little bit more details on the stark friendly and there's not really hash functions so there's been one of the candidate hash functions in the competition that was broken the GMC so there's kind of two of the more experimental ones that remain rescue and facade on but kind of the bad news is that the the Crypt analyst in

the in in the committee so there was a committee that reviewed these hash functions they still think that it will take years for these functions to be very far away vetted and for for us to have very high confidence so actually won one of the hash functions which we do have available is the the pieces and hash which is kind of provably secure under well under trusted trust assumptions so one of the things that would also be worthwhile is looking at the performances both in the plaintext

because it is significantly slower than judge 56 but also also in the context of another alternatives you were kind of

Vitalik Buterin: hashing backstops by the way it is kind of bring back template multi hashing so basically like at least for shard blocks they would provide both the regular hash Rouge and the MC hash root or whatever hash function route that we have and then you would have a fraud proof to show like basically that one doesn't actually doesn't actually lead to the other and and so that way we can kind of

benefit from the data availability checking keep the conditional and that has that hash function working but we also have the medic hash chains that are backed by whatever

science we have t6 right oh but I'm Todd wasn't wasn't totally convinced that this is necessarily kind of strict security upgrades I guess discussion I guess like what what is the are the arguments I'm wondering well basically he's I think

he's

Justin Drake: saying that if the hash function is broken then basically you can trick

like clients and

Vitalik Buterin: you won't be able to trick white clients because white clients would ask for branches that goes through with the routes that go through the traditional hash like you know if the hash function is broken the only thing you'd be able to do I think is get away with unavailable data if you can get a committee to sign off on it

right so if basically

Justin Drake: if the committee is are compromised and the hash function

is compromised

and it's yeah okay let's keep moving anything else racing will hey what's up everyone um so we're starting a phase two community called December 3rd so I've been an issue in the and the PM repo and recently released a update on everything up until this point so you guys are all interested in progress and what's been

going on definitely recommend you you read through the article we had an implementation this this last week of basically a runtime api so this is the sandboxing api for calling calling new wise um runtimes and those functions to support this and run them in a a protective manner so now we're prototyping a smart contract execution environment and we'll go ahead and get some basic smart contracts running for

that we've been expecting scout to test across multiple shards so this is this is perfect in tandem with the smart contract yi and we'll be able to start testing contracts that that are sharda where we're continuing to optimize the toes knees and again and the the update we I think it talked about like six token EE implementations that were going on in in parallel and also continuing to work on making the some of the accumulator implementations more more performant performance as well efficient

for you we

Mamy Ratsimbazafy: what language do you use for the smart contract so specifically the smart contract that will be called will be in in Wazza right is that what you're talking about

yeah

Mamy Ratsimbazafy: it's because we have a proof of concept so with an int wasn't - right smart contracts but we are looking for new targets like what VN we can use it for if we can use it for a freedom - it will be very nice awesome yeah that

that'll be cool we'll play with that I think for the most part the contracts will write will be put together an assembly script and or rest those are those are what we're playing with initially and then having those compiled that in

tourism yeah is that answer yes

Mamy Ratsimbazafy: if we can find the contract original one we're a source on this script we can just replicate it in name see that we generate something executable that is almost the same on this way we kind of well it would be a bit like Viper in terms of interface and that we can use it for fish - for

testing also awesome yeah let's do that so all I'll share with you is we get this get this working and then we should try and not we should try and write one as well

and that yeah cool all right thanks will other research updates okay moving on

Danny Ryan: I want to announce them let's shall be aware of there are some new domain specific calls

I think the particular cadence of these calls is still TBD likely once a month to start but might depending on progress and things might might swap to be quicker so there's the East one X there's a research call they're primarily organizing on the e3 search forum so if you're interested check that out it's pretty much a provided revitalization of an effort to push forward on some of the research on the existing etherion chain with a focus on getting one improving theorem today and also staging ethereum for the future

upgrade of moving hearing into to moving more towards stateless model and other things so if you're interested check that out like I said they're organizing a lot on on the research forum we'll announce a soup oh I drop the link of the chat and in addition to that we're going to start and at least monthly networking call where a member or two from each team I hope that it has been focused primarily on networking we're going to gather gather enumerate the problems more explicitly and work on driving that effort a little bit more to

sleep I'm going to leave that call to start but certainly I'm open to not leaving that call depending on who steps up and is available I will drop a an agenda item on the 8th - p.m. repo an issue that will set up this call it might be next week but it's Thanksgiving in the US so if not it will certainly be the following any questions or thoughts on that particular item okay and obviously don't want to induce an

incredible amount of coordination overhead a for adding all these calls if there they are experiments and if they're valuable will continue doing

great Danny one more call yeah oh yes absolutely I'm sorry caiman yes the as I dropped in the secret birthday yesterday there is the light client task force which has another call coming up and what the first week of December that's right yeah when Wednesday so we're gonna be twenty twenty three hours before this next call so the next piece

to implementors call great thank you and I mean as these two progresses these two covers an increasingly large domain and increasingly large number Chris pants so hopefully with these domain-specific calls we can have rally the effort and capture the duplicity okay moving on to networking Felix this year how's it going Felix any updates on your end yeah no unfortunately there there are no updates since the last two weeks because I was really busy with my cat related tasks one thing that might

be of interest to the historic team is that we just converted key going to use four modules which means it's actually much easier to it should be much easier to actually consume Goethe immersive tenants now and I'm looking to include the disc if I code on the quartermaster branch as soon as possible so I think that's gonna be a big help to anyone using gold because then you can just report it there are key from the upstream still existing they're gone and then apart from that what we also like this not directly

related to each other it might become like interesting tweak to in the features since the Istanbul heart fathers coming up on the each one chain everyone is upgrading their notes which means that we'll all of a sudden have a lot more ian or capable each one notes around and we want to leverage that by building a like running a DHT crawler that collects all of these universe and makes and the code for this is done like we have the crawler and we have like all

of the infrastructure to make easily right now we're just waiting for people to base your break so you can actually set us up and then I guess we'll have something to announce there right after the after the for process and I think this is gonna be a pretty good test run of this like DNS infrastructure and I think is probably gonna be really nice for it to as well once it launches possibly on to c5

Danny Ryan: I know there was talk about integration of it with work or looking into other techniques for adding some salt resistant resistance is there

any update

on that no so I haven't received like any feedback on on on respect in the last two weeks this is I'm not sure maybe anyone could come and look I mean I hear a lot of update on this call on day two call and I'm really happy to hear the people actually implementing it it's just that like I'm not really in contact with anyone so it just all seems to like work without my involvement which is very nice but at the same time there are research challenges left for this protocol and I also be really happy to like get more help on that like

shouldn't just be me I think so yeah I mean it's nice that everyone is implementing and I'm really happy about that I can really say that but it's just like I would be nice it's like there was more involvement in me on the open research questions there it's not there are not that many questions left actually and they're all on the on the track first one I'm also happy to have calls about the city anytime yeah cool

Danny Ryan: Thank You Felix networking experts provide feedback and input and like I said we can have a networking call

coming up and we came out more explicitly rally around some of these problems yeah Thank You Felix other networking updates the on the harmony simulations when are y'all expect him to publish some results from that

Mikhail Kalinin: I think the basic results should be published like the end of next week I mean the results like we're just going

to publish this basic simulator stuff get feedback from protocol apps from my proto who is willing to do this to the creation simulation

yeah but move forward quite cool thank you other networking updates okay spec discussion proto dropped a issue yesterday evening or whenever that

was discussing some of the pain points with the disparity between hash tree roots and signing roots proto would you like to talk about that yes it's okay so signing root is kind of this pain because there's all these multiple different domains there's the last client work networking this flexibility of being able to convert signing roots or a route of just a message into a full history route that includes the cincher

and then there's tree based state representation it's like kind of closely connected lifelines work but also if as the equation work and reduction of state representation so you can store the state in some kind of miracle store to serve light client so for example all these kinds of things they're blogs are very inconvenient because of the way you have these two different kinds of routes let map to different tree structures and a outlined this in more detail in the issue and I should are even more like edge cases and things to consider here

select a more feedback about this approach so to remove sending routes or alternatively there are other were runs but they're not as pretty and don't cover all issues but I think some feedback here is really important as it's close to something that's frozen itself is not quite frozen so I do think it's worth the discussion and hope to be like with more feedback I'll put together a PR describing the changes and then by the next call I'd like to make a discussion a decision as a group of implementers on what we should do and

if I should remove it or if there's better ways to do with it thanks Fred oh do you all have any questions or comments for him other than just digging into the issue that he posted hey also when I add on to this row from prismatic labs here sending Roo there's a name is also can be a little bit confusing like to somebody that's not deeply familiar with its purpose or the EQ spec so yeah I agree that it's worth revisiting also it doesn't make sense that only something like to an external implementer like why sometimes some types types tiny rude and some others don't so there's a lot that can be done to explore alternatives

I like photos alternative of having like a sign block container and like also regular block container and he mentioned the point that it could just really remove remove a lot of complexity that exists regarding like dealing with latest block headers and things like that in the state so that could be very helpful intro I know there are a lot of benefits to be had for just a little bit more for ghosty ghosty in the spec and I do think even that can be really minimalist and it really just needs one Ripper type to

say hater is this sign person of the container there's the danger and that's pretty much it actually that's

Mamy Ratsimbazafy: something that we wanted to do in English one suspect stabilizes because we have issues with on the test suit with tests that are using do mean signatures we we need some kind of workaround or undocked I guess having a sign type and somewhat something separate it will be much easier

icon for testing right and then there is something to networking as well

Jessica constable and then there are considerations for how you deal with typing structure as a surfer for life plants as a live client to consumer proof take a look and then we'll discuss more next on thank you thanks Beth other spec related items let's talk about okay I have the next item in there was

general discussion of test nuts I think the status is very

Danny Ryan: similar to two weeks ago there is a the prime focus for most clients is getting public versions of the test nuts out and or joining other public test nuts there's obviously some really good movement on that with prismatic relaunched in their test net and with I know a couple of other clients very near that oh and Nimbus doing their weekly builds but it seems like we need there's still work to do before we orchestrate a larger scale

multi-client public effort are there any other thoughts comments discussion questions about this

this week great okay the last item open discussion and closing remarks anything anyone wants to get up

Mikhail Kalinin: I have a question about weak subjectivity period weak subjectivity

period size comes from a Casper FFG paper and I'm wondering if we are like tightly coupled with the size calculation that we have so far and does it that be to prevent long-range attacks only or there is some other implications here

the

Vitalik Buterin: weak subjectivity period yes so then there are calculations I made Annie's research post about that so basically how long the weak subjectivity period is based on like the rate at which people can withdraw and the

current the withdrawal period is right now it's a maximum of 8 months in the worst case like in the normal case when all of us the one who is what's drawing you get out after about two days and then in the case where there is a small amount of ether than the maximum also drops that's right there and of compromise and encourage more people more people to join in and in terms of why that exists and based one part of it is because it determines like how often people need to need to come online to maintain to get the security guarantee year another aspect is that it Creed determined like basically if the weak subjectivity

period gets too short than there are made then people who did bad things might end up not getting slashed so it's both of those things put together

Vitalik

Danny Ryan: do usually define the period as the time in which it takes

and so I think that our whole

Vitalik Buterin: I think mathematically speaking it would have to be the period that it takes for a third to exit

right yeah I need area studies research post let me see I'm gonna go searching subjectivity it was drawled you wrote

about this okay so there's number four nine four two but no that's that one's like not detailed enough it's called weed subjectivity under the exited if you model five one eight seven okay thanks with Alec

Mikhail Kalinin: I'll take a look at this I was just wondering if we can cut it down by half for example and

whether it just breaks long range attack defense or breaks something else

Vitalik Buterin: to break what down by half the maximum or the amounts

so yes there's the there's the maximum and then there's the lead of four per epoch I think yes but then they're they're really like the effective week

Danny Ryan: subjectivity period is the period in which the network is actually policing right which could be up to the maximum theoretical or below

yeah no

Vitalik Buterin: I guess the maximum okay so the

maximum withdrawal period is theoretically eight months but that when we happen if literally everyone tries to exit at the same time and then the amount that you would need to log on

would basically be like once every just eight months divided by three it's many we need to get slash that's yeah and that's also the word and then the eight months is the worst case and like in the average need to wait much less and we did some sub simulate this

right okay cool cool

Danny Ryan: other questions regarding this or just anything during our last open discussion period okay I know y'all are all pretty heads down working on these things I think this we're also entering into a period where it's very important to communicate with the community so that I

started doing those quick updates but updates and your individual teams on how to better understand your clients again vulva test nets are always extremely appreciated by the community so don't forget to write thank you keep your eyes peeled on the eath PM repo I will drop details about that first networking call and I believe we will do this call in two weeks I'll post an agenda for that soon thank you everyone

talk to you soon [Music]

[Music] [Music]

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