All Core Devs - Consensus #35

Transcript

[Music] [Music]

[Music] [Music] [Music]

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

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

[Music] [Music] [Music] alright here's the agenda and probably

have it share it with the chat chat on YouTube let me know okay most of you probably saw we really cz row 11 last night this includes a lot of dose hardening on the network PP spec as well as some other changes that have been introduced if you'd be set to clean things up as people have been working on their test stunts and just that there's a few things that made it into the same transition spec including a little bit of extra support for isolating more

easily isolating chains adding the proposer indexed blocks to kind of increase the usefulness of being able to verify things if you don't necessarily have access to all of the state and a couple other little things this is to be stable once a client target this is something where we are not intending to introduce change in state transition I do expect over the next few days as y'all implement this stuff to have some feedback so some clarifications and maybe some little fixes to make it in and from there I as we are doing all to climate testing and see potential

disparities and how people interpreted things or better ways to do things I do expect to see a little bit of changes to make it still into the networking spec especially certainly clarifications so that's that the next testing priorities are drumroll for choice and some of these networking tests that proto has been proto Dmitri written some stuff up on now that a lot of the burden of some aspects stuff is also off of us I think we made some progress here questions thoughts

hurt or anything yeah regarding testing updates better lighten up her sorority has been a pleasant this apart server done is mostly in preparation of the new version it doesn't change all that much just a transition we needed his view us and he needed to verify all these test results us internally the spec has been changing and getting faster so this factors are out for new person as well and then it's more about network testing integration testing from now on and as you know we

work on this new tool called reamer and the rumor Atlantic in that room are into the scripting approach where we can set up test experience mention bunch of nodes asking for different block requests in the client wisdom name with the blocks so that the basic same test these kind of things they should be really easy to script get this name setup and that's notthat's thank you other testing items

great client updates I want to hear how things are going and also if you've done any multi client experimentation upon discovery or actually doing something stuff we'd love to hear about that how that's going let start with Sigma prime Plato's hello everyone Paul here so we've started some work and research on a peer reputation system we've also pulled out you know out into its own great so you can share it with open aetherium or improving a defied implantation to work with the new in our

crate and also to all our lighthouse to make some performance tweaks on the mechanism you know for like I've been limiting modifying timeouts and things like this after chatting with prison who Nishant has been doing some great work doing interrupt stuff so I was chatting with them and protofelix we're improving this 55 to limit added peers and a DHT so we can speed up peer discovery we're building an en our predicate query search and we're doing all this stuff so that we can do our upgrade to use the excitation aggregation scheme it's in progress but coming soon we also release to validate a UX survey to help

understand what's takers need one saya that's wrapping up soon if you're interested in contributing you should be able to find the links throughout Twitter we trial as a light implementation of proto is remarkable and with some good results but we're not quite moving across to it yet because it's about a significant change and it's been didn't quite quite more than quite worth it at this stage promising and we're also experimenting with some optimizations on Baylor signature verification straight up box processing and we're seeing pretty good results so far over the past few weeks we've on-boarded two new developers we have Adam and Eva and they've been great we also released a blog post this week if

you want a little bit more detail that's it from us great thanks Paul prism hey guys um Terry's here so like Paul said we were able to interrupt with my house on their EP ever branch that's super exciting we were able success we seemed to head and the prism didn't have any validators so prison which is basically busy and sinking for the still really could you see that and our version 0.1 of the slashing service is at hip on the prism test net so we verified that with recent sense to be called a double boat and then there was

also reported here become flock explorers and then the validators was indeed slash and and ejected and now we have this new SS z4 prism which is significantly faster than before so this was achieved by creating a code generator that that outputs more emotional functions for Amy's front window that is valid under the SSE rules our OSS he was pretty slow because coders don't have support for generics and then now is just a time brush so once this means SS V is used in prison we will announce it and then I'm also

working on station service which allows generation of any of our fishery is to grow state so that should significantly improve on the memory usage and then the CPU usage as well there are a few items we're creating working on so working on long dynamic substitutions of subnets and they were also working on no no he's in durability and at last world also what we are upgrading to the latest spec release version eval and that's it thanks cool thanks Nimbus hi so we created all new targets for

March sprint and as three highlights we have one interrupt to starting the auditing process so that we can at least discuss with firms especially given that name is a very new language that we're probably not know and further highlight will be dropping completely to be given in terms of a date on the course pack we are out more almost at the open 10.1 with the new BLS changes merging has been delayed due to low-level bugs and we have progress on the attestation

aggregation work it's not fully done but it's malleable on the testing front now you see I automatically launched a small local network and ensure that we are finalization after five epochs on the networking front for discovery we can now interrupt with Jeff disc v5 we cannot yet interrupt with light house but we are testing and making sure that it will work we have several improvement to Lippe to be stability and also we have a new log analysis tool to track to the errors on the performance side I

mentioned last call that we had some performance issue on on the single don't and we actually found out what the bug was it was a bug in the p2p where we had a buffer of size one instead of one megabytes what was just a shift right instead of shift left in the code and so after fixing this bug we know can handle for five more five times more traffic but the downside is that we generate five more five times more logs but it's a very nice performance fix and that's it for me thank you very much

Tikku okay that's me this week so in terms of progress discovery v5 is now merge game modulo a boggle - we can discover modes on the lighthouse Network but we can't discover our own notes I believe so something is going on there a deposit process processing is all done sync seems to be in good shape now and we're making very good progress on RPC implementation as a heads-up on that there's an interest within Pegasus on working to standardize RPC interfaces across clients already there's a bit of

divergence between clients there was an effort that started a while ago in the github repo but this seems to be on the back burner at the moment that's strong interest in this from in fira as you can imagine varying RPC interfaces are a major pain point for them so they may get involved with this or lead an effort or something like that so just wanted to give you a heads up look out from beneath more news on that on performance we've finally merged in a massive PR from Anton to implement remarquable we see four times speed-up in block processing so we're down to 200 milliseconds and a four times slow down in epoch processing twelve hundred

milliseconds is a win overall but we're now working on improving the epoch processing 98% of block processing time is our signature verification so that's the next target will likely move to Harumi and put in the load of caching and other other optimizations a big chunk of our sync time we believe is database operations so we'll also be working on that next sprint and finally we are ready to merge zero point ten point one into our master branch we've been delaying this a while as we've been testing our sync against president sapphire Network but we think the time is right to move on on the same

day that the zero point eleven years that's it for me cool exciting progress the asymmetry in the discoveries pretty funny for the signature processing I I think I'm pretty sure that lighthouse is the only team that's implemented this fast verification of all see the LS signatures this post from Metallica while ago on e3 search I think there's probably on normal blocks like 20 percent ish games to be had there so teams you should take a look at that link I just sent otherwise artist sounds great thank you let's do trendy everyone

let's see no crazy updates in terms of features except for some stability of the overall node and are lucky to piece back we do have some exciting interrupt news with respect to disc v5 we're interrupting with I believe death and then definitely lighthouse so that was exciting to see and otherwise yeah working on updates to the latest spec release he's Alex and then yes I'm also keen on the API standardization it has certainly been on the back burner with everything else but some extra eyes to see five past few weeks it isn't quite

ready but I think in the next week or so we'll be ready to start testing it out interrupting with others we've been thinking doing some work we've been doing some work on storing storing our data more efficiently to the database this has been kind of an issue as we've been storing every state and we're working to fix that we we just got some results back from some audits we were able to audit some pieces of code that we don't think are gonna change we're very unlikely to change so in particular I P 2 3 3 3 through to 335

like BLS key generation and storing the disk and the results the audit came back fairly good we should make them public here shortly once we fixed just their their minor suggestions and in the next few weeks were we're going to be trying to start syncing from other public test nets and just generally trying to improve the stability of our of our node great Thanksgiving and to us perfect timing never mind all the updates from sly

coming from last week so there was a two point ones back and Tobias the last requirements version of T of the single commissioning AP a nice and steady splitting of the validators processors to be able to run the test so we have the two separate processes synchronizing together so significant progress a lot of work from sly who are super excited about the achievement in the last two weeks we are I'm - great thank you very much

I believe that is everyone correct okay next on the research a couple things I wanted to get in the we do have the audit results of the Yui story audit which would be to release ASAP some of those changes already did make it into the spec otherwise they're kind of some suggestions for future work which certainly do attention also phase one is and has been getting a lot of love it's kind of some testing in place in a major PR to be merged very soon with a few cleanups at which point at

least the the general structures are there probably be great timing starting last week of March May 1st of April if anyone has an extra resource which I'm sure you do to begin prototyping that might be a valuable kind of feedback loop there great so research updates let's get us started I'll go bertie exotics so we had to write ups come out since our last call one is covering the

safety of the each one follow distance and that's written by Mick il because he's working on research into each one finality gadget and a two-way bridge we also had another write up come out covering non sequential receipt cross sharp transactions and that's just kind of like an appraisal of that and that's it I'll leave some links and chat Johnny's not dead and that's cool Johnny had surgery right is that what you're saying sorry yeah yeah yeah

kidney stones then he had surgery so it's not dead good that's always good side here yeah I was thinking thanks Joseph Vitalik do anything you want to update us on yeah I mean I can talk about the post for comment that I made a couple of days ago about using a polynomial commitment stop eight state routes so basically the idea here it's this more well longer-term proposal that basically says that instead of using the more binary Merkle tree is there any kind of Merkle tree to store state routes and

this could presumably apply to the eath one state route this could apply to states of execution environments or other things we wouldn't use polynomial commitments which are this and if mathematical object said well that's that's kind of a hash of a polynomial that that you can do certain kinds of mathematical checks on and I talked about how if we switch to this kind of model then it becomes very easy to have a very short witnesses for a very large number of these endowments that you're proving and so to me and a current bottleneck of stateless clients which is basically witness size would pretty much

dropped by potential yourself something like over ninety percent now in terms of and of what this actually requires this basically requires this kind of structure that I talked about in that III search post it's it's called I just am using polynomial commitments to replace state routes that's right near the top where we have this kind of scheme based on several layers of caches and where you have these kind of multi multi proofs going um that prove a whole bunch of values in a polynomial commitment at the same time and you know it does rely on some some

kind of very new math like you're requires it relies on some stuff that ask tech people invented literally two weeks ago but everything seems to and of everything seems to work note in terms of Ana practical implement ability I think in the short term the Moriya the most realistic place to start using polynomial commitments would not be state route so much as it would be block bodies and I mean potential potentially receipts as well basically because with the block bodies that receipts you don't have to do any kind of proofs that show

that I commitment has been updated correctly or anything like that you just have one commitment and then if you wants to access that you can just go and access it and the shell and that's something that I don't know I personally think that we should they potentially even consider using this instead of Merkle trees to store things like shard block data fairly early but it's still something that's just worth considering at this point and some of the benefits would include the smaller proofs also you get a racial coding for an of data availability checking basically for free

in a much simpler way that doesn't require any kind of fraud groups and you get for free I mean potentially better proof of custody schemes so that's something that we want and if think think about more as well and then in the long in the longer term this is something that could be used for the state as well so you know in encourage reading the post and encourage and of leaving questions thank you and the pass we will take for the time being is to make sure that these types of components or well black box

such that it's time to switch over we can easily swap it one important kind of like boxing thing that's important to keep in mind is that these schemes will end up ahead of changing how the um how witnesses work and so like we don't want so for example if someone ends up writing a merkel branch verifier in the EVM and then later on we switch from more coal branches to okayed commitments or to fry then that would add up breaking and so it's yeah I guess importance am i important developers or again the protocol such and may not make

those kinds of mistakes don't lock us into things thank you do research updates and one thing that's worth just saying is that the conversations I have we had with a snack people in general made us kind of quickly much more optimistic about the medium term possibility of doing this ek snarking over a full virtual machine so that's something that's at least in my mind kind of moved

from crazy to an ER medium to long-term plan actual possibility right which could have layer one uses and certainly would have layer to roll uses yes I could give a quick update from the quote side so actually there isn't all the other much stuff it's fun because we've been mainly busy like going to each CC and it's one one exam it was good great to see the progress and on the 1x item damn we're now looking a little bit into

like how or if this is this thing needs to research that might be useful in the each WiMAX context so for example like that might be state provider network Sam has been really interested in the DHT approach there also things like SSA static state exist which I've been talking about last call how that might be useful in the state of mind in context and stuff like that we're also continuing our research on these two side of course but I assume we might make sense yourself updates they're on the next call because we hadn't oh it didn't have all that much time to focus on that yet I think that's that's that's all already from the cold side Thanks

other research updates okay thank you everyone next on the item I'm like homeless is networking we had a what I thought was pretty productive network can call about a week ago there are some notes floating around from then good and we've seen a few kind of items shake out from that I haven't decided if we're gonna do one next week are there any thoughts on that

are there enough networking items or do we want to kind of get into the spec and then do that two weeks from now anyone feel strongly about that otherwise I'm just make a decision later today okay any networking items we want to address today well just announced so during if you see and after if you see I think I'm working with Demetri and Felix Dimitri from our status so letters are the discussions the sting has moved

for they have approached now and like ideas how they're going to make this work and also the help of Felix the governor pick me discography five has some minor problems these fixing these and actively testing against David clients so if your discography five is ready or if it's like ready to like get dusted against other clients or at least gonna be to be vertical if area is completely flaccid mutation then please do share your implementation build

instructions and local to your testing other networking items okay anything spec wise before discuss cool I know you're beginning to take a

look at easier 11 I know there's already issue on spectra though you have a lot of conversation there okay so I think the way we are framing things right now is marches multi-client experimentation and giving up to v-0 11 whereas April will see multiplying test sets obviously that is contingent upon a successful March but questions or thoughts on that great open discussion anything you want

to talk about I've actually been wanting to do another sort of client retreat inter-app thing and my on rotavirus and it looks like we're not getting planning that anytime soon this for the next but it's a lot of good work good collecting a half a million turn up anything else and I we all with a lot of work to do these fees one implementation like do we

have a roadmap for even one implementation of phase one starting anytime soon there are two or three teams that I think intend to put a resource on this kind of when we signal to do so and signaling to do so I think would appropriately be proudly end of March first week of April given the status of those specs and I hear whispers of eath one client teams wanting to prototypes an eighth one integration work against phase one prototypes so see that movement as well in the coming couple of

months okay okay okay these calls are easy when there's just a ton of work to do on y'all's end anything else okay I'll let y'all know as soon as possible when the networking causing the Emily's call again in two weeks thank you everyone secure [Music]

[Music] [Music]