All Core Devs - Execution #42

Transcript

[Music] [Music]

[Music] [Music] [Music]

[Music] [Music] [Music] you [Music] check-check

okay yeah the audio is working hoorah I think we're gonna great hi everybody welcome I'm gonna pull up the agenda but we were just talking about he was them and basically what was the question someone had like should we implement it before sharding and all that stuff yeah like should he wasn't be available in Maine that somehow so is anyone from uoit's I'm here to like kind of give an opinion I could if you can hear me yeah what's up it's Alex or exit yeah just reflect to the questions from Peter we are working

on a test at a Curie was on test net and Peter mentioned multi clients port in depth so for that we have CPP TM is our main client at the moment but there is work going on to support material which i think is fairly it's close to completion at this point and so the test is hopefully gonna run both CPP theorem and to its hiriam and then as regard to whether it could be introduced on the main net before that we have a proposal in progress which would offer a subset of D was to be considered for the mainland and that subset could be used

to write pre compounds while things which are written speaking parts today to write them as was in contract but it would still because it's a subset if they still require consensus to decide and but if we take them onto the network or not could be like an intermediate step and I guess we could even achieve kind of games and simplicity because we could basically force all the existing freaking piles to be river isn't a specific pieces of Watson code yes correct we we actually have those rewritten the the benefit of this from the wasn't side would be that it's a restricted set so we may not need to

address some of the concerns which have arisen around wazzle and we could still research them because the these set up recompiles which would be combined to Wasson would have to be reviewed by other older client developers that need to be consensus around them there is no risk of random this is nice because then we can make Zuko happy about Blake - we can make the elements and so on and

cohab i adding finite field operations so you can use our purification a bunch of things yep like en SP compiles which ones are those again oh you mean like there in the weird hash functions that are in signature schemes that people because like the traditional DNS system uses them yeah yeah there's one other one that people curve to five five one nine I think people wanted so this is proposal is currently being worked on

because there are a couple of different options to be explored and one of the options is whether it should have its own wintering rules and if it does it would have to be only written it wasn't because it's specific to a zone but if you decide to you have this random metering logic we have followed so far for compile and it would be possible to for clients to decide that'll be able to use this single written in wasn't or they could just keep writing their own version of baking pies but we haven't reached decision which one makes more sense and wanted to get a bit more detail on these before we

we proposed this to this whole but I think we really hope to to have like a proper proposal like two posts from there to be discuss cool that sounds good if anyone on youtubes hearing an echo it's because I'm in a really weird conference room where I'm not familiar with the settings so sorry in advance I'm just gonna try to mute when we're not talking cool so I guess what came from that is you said in two more meetings from now acts if you're gonna have something if you already have an issue be put to discuss this and everybody is welcome to check it out and yeah I mean personally I

would be happy to discuss this even on the next polar the collator but we wanted to get some more detail from our side before we propose it but yeah I don't mind better where do we keep discussing it on the next ball hmm okay sweet let's get to the agenda unless anyone had anything else with that cool first off is Demetri here for testing okay we'll skip testing for now then let's just run through client updates real quick I guess I know we don't really have any specific updates last week

during the to meet up policies like tying polishes we're trying to figure out a way like times to be able to sync to the change faster etc so we're mostly hiking a lifetime stuff and no dilemma August cool parody we are planning a major release very soon okay sounds good at Harmony we have finished

some current activities this weekend starting to work become training since Monday so yeah that's it pretty excited about going on research cool Trinity next major release [Music] architecture that we're working on for plugins to be able to host a third party functionality that kind of excited about but in the very early stages but generally working on synching logic it's

sort of still a lot of the core functionality cool C++ aetherium Paul you have a CPP aetherium update or I think you're muted if you're trying to talk okay we could skip for now Pegasus what's up with you guys on here I'm here with Ben you all know and Matt Halpern

one of the quartet's working on Pantheon thanks for having us to talk about Pantheon today I'm very excited to announce that Pegasus is planning on releasing our Molinet compatible with Pantheon at DEFCON 4 so some of the details around it are that the client will meet the requirements of a well behavior here so it'll be able to synchronize the entire chain through all the hard Forks mind blocks and propagate pending transactions the client will also be developer for friendly supporting common app development like truffle and web 3 and deployment news cases this also means that the client will support JSON RPC ap is used by the

popular clients such as at net debug admin but well yet have technologies such as whisper or swarm as we're focusing on core functionality robustness and modularity we'll be providing more information and insight with what where we're at in the coming weeks and we're also planning on being more active in the community such as providing feedback on the attorney reference tests on running web programming programming language written Pantani's written in java yeah guys

relations yeah congratulations thank you thanks for listen yeah sure thing cool do we have anyone from Nimbus or mana or any other things turbo gas one of the last things to test we work everything again [Music] even the database like I'm using now

is a very supportable to what we actually need in turn I mean I've gone to some database research and I found some data structures I'm not going to go into the details here yeah so I'm getting close great update from a wasum you kind of gave one earlier axe ik but if you have anything else to add there's too much else we still plan to to release this test net but we don't have a final date on that yet it's hopefully

is good for Def Con anyway the doctor the evolvement open challenge if anyone out there was right and view our token in their language code it's true Daniel thanks for reminding me have a subset of a 20 token feature subset of ownership 20 token which can be implemented in any language compiling to Lhasa and they're all ready for implementations one can see one handwritten one in a subscript M a rest version so anybody is sp2 to write any language of choice in them and

we can test all of those nice cool I [Music] hope no one's using that if you're out there on the internet and using that stop it there's better things all right so that's cool that it's good a that's that's good that we're getting it officially deprecated thanks Martin [Music]

that's funny okay so we'll get to research updates now oh yeah go ahead Peter from the missus team here yeah um I didn't I didn't see anybody I didn't see anyone from aetherium j/s either is there anyone from mist here anyway so it's not just time can you update and I'm not sure whether they don't you guys that they did a release a while ago which basically when you start of mystical tomatoes it basically connect students work when it starts up and then switches over to the full note annotations thinking so that should be a really really nice update or anything

using [Music] weather but they told this an important meeting or not that is not done here we go oh yeah they haven't mentioned it on a for deb meeting but I did see they had a release like a week or two ago about that it was like right after their last for deaf meeting so that's really cool yeah now we're actually getting fuck-fuck reports that when people start up everything works and they see their banisters and when they stock up yet they don't see the madness nice um okay so for research the talaq do you want a leader do you want Danny to start with ornate or anyone else well Vlad Nate and I have been arguing about

chain for service rules a bit basically trying you know the gem the place where we stopped as like just like this is already getting into optimization territory but trying to figure out like what are these issues where well Nick sure does sorry just to kind of dial back a bed mobility we look Justin and I am at Burley and started thinking about one issue which is that like given that using like pretty much every random beacon scheme has its own flaws of different kinds so what look for example ran'tao has like exploitability the threshold VRS and

have this like hard threshold online requirement and VDS and have to worry about and they have a high chance of being monopolized in all those things so we're basically trying to see go back a bit and see can we make the algorithm like the the beacon chain algorithm itself more resistant to like a high level of a manipulation of proposal selection the basic arguments here is that because the a tester sets are very large it's just basically statistically impractical to even if you have perfect control over the random see it is statistically infeasible to really manipulate like more um even one a

tester set but the number of proposals in one epoch is like 100 so that's totally manipulable and if you have if you can perfectly manipulate the randomness so the issue the thing that we're trying to figure out is can we modify the algorithm to be kind of much more robust against proposers and being really unreliable and the security assumption is that like worst-case imagine that within one epoch there's only like one or two proposals that are honest and we think that we can but it basically requires some fortress rule modifications so the problem with the old fortress rule was that the old fortress rule was very kind of proposer centric and that accounted by Heights

and proposers have this large power to basically prevent Heights from incrementing or false Heights to increment and the scheme that we're looking at basically move either well one option that I like I'm thinking about is moving fully to something goes to like and one option is moving out of partial weight is something that partially in the ghost direction the idea being basically that these kind of like base if you have like chain a and then on chain a proposers just stop showing up then you have all of these dangling at the stations and can we make the dangling at the station still contribute to that I like you can to the head score so that even if you

have missing proposers for a very long distance and then some other chain starts growing that other chain still can't overtake the original unless it has more than 50% and then at the same time also trying to see how that fits into media finality gadget and budget justification and I kind of all of all of those details so making progress but still wants to think a bit more and also thinking about knowing making like the design space between sort of FFG style algorithm to see this new style algorithms and what the differences are and what the intermediate points are and so forth cool any other research updates or comments oh yeah we also have a guru

I'll go ahead I meant for the call but we're gonna do it like I don't recall every nice cool anyone else so we also have some questions on the agenda so basically because the IP 1011 the hybrid caste perspect was now deprecated is there still interest to have a joint test net like a Rob Stinson test some of this new stuff maybe ewaz um and other things that's the first question the other ones are if there's an interest in this effort how would it

looked like if there's not a hybrid Kasper proposal and what stem should we take to in the meantime while Casper and sharding is still being developed so the first question is basically yeah what should we have a joint test that Wow okay let me ask a different question joint that's that's for Casper let me see how I made that comment I feel like if it's in the ultra short term like a couple of months and it should be a joint test that just for was um if it's some longer-term like if it's longer-term to the point where the beacon change details were solidified then a joint test set for the beacon cherry would be really cool

yeah I think if you want to this [Music] [Music]

particular but feature-wise Casper we don't have a gas shortage to the test program currently first of all we could distinguish between Casper and Casper was shorting because like the beacon chain for the Casper was shorting thing can totally by itself be used as a

single as a single chain Casper set up that thing using the hamster contractor right I'm not suggesting about that okay so yeah it sounds like what there's not gonna be a shared Casper test net until we get much further down the line so the next the successor to Rob Stinson look II was him but I guess my point is that maintaining the consensus engine is gonna work because small quirks and so it's completely fine to do it if you if it have to the fly mass effect but it's just a temporary thing

yeah and it sounds like it's more of a temporary thing cool so what was the other questions on there I think there was like so it sounds like we are interested in an effort for a successor to rob Stan it sounds like if we were to do that we want kind of something like he wasn't to be put in there and I know that there's being an e wasn't s net developed right now is that is that basically exactly like the main chain that with wasn't plugged in or is it different that one currently has two courts that bother me personally one of the perks is that they removed the code size limit which is probably it was on contract require a bit larger fruit sizes but

that's that's a bit problematic point because well essentially that's also denial of service attack back there on the transaction pool so it's we can't just allow arbitrary large transactions okay sorry can I so it's not only the code size limit but its together with another change where we don't discount zero bytes in transaction data anymore yeah the other one I don't really understand why so basically network traffic is compressed and the storage on this is compressed

meaning that unless you have alternating zero one zero one it should be able to compress it fairly well so why why did you guys decide to change the pricing thank you I mean the zero bytes was also one of the key attack vectors because it doesn't matter it's compressed I mean that's an implementation detail that were it's compressed or not but you definitely can have an alternation of zero ones or donation of anything and zero which cannot be compressed anymore right well no it still can be compressed because like if a zero makes up 50% of the compression like in compression

algorithm that's remotely close to optimal will assign zero to the zero byte to be like less than one bit so the like you should not be able to make something whose compressed value is like more than twice as long for per gas than just random bytes I either way so the exact changes are probably be late to go and what's problem data but the right time currency to debate but I wanted to make is that if you want to be a scientist then I think it would be nice to somehow try to keep the changes the minimum because any kind of soap changes that don't really make sense until you bring too much

value they always pose the risk that somehow they end up being enabled on mania for example they do that the parity had drops and that they accidentally leaked out the IP there so that's why I'm afraid of all these tiny tweaks that don't necessarily in banner okay so it sounds like we still need more discussion and kind of hashing out of whether or not we're gonna have another test net successor that involves it wasn't or not I think I have just does that make sure that I believe more the change in our achievements great so yeah

what were you saying Danny Oh a joint that's not ready to go when we want to do one that doesn't have yeah we can just keep Rob's tempore that or we can make a new one yeah as Rob Stan does anyone know the current state of Rob's turn really it's kind of been flaky lately it's kind of like a stepchild I guess thirty specif is mostly koban yes that is mostly wrinkled baby yeah

what is the minimum sort of client support that we need for a class time like woody blossom at this particular moment we only know well I guess the point I personally would at least like to see both Joey theorem and heretic to agree on respect simply because if if we introduced some really new yearly idea even without modification that is really hard to integrate when the other one we run the risk of not having the other time so it would be nice to at least try to target the mature to try on at the same time

so actually I can regarding the test attention on e wasn't we have a single VM in C++ which works directly with CP p terraeum and there's also work to which is almost finished to have it running it kuwaiti are you but it's still a single VM and next to it there is a separate report to have a native go interpreter for was and I think it's also like it's in a prototype stage but it's kind of working and one problem though with the test that we are facing is CPP Tyrion doesn't support proof authority it only

supports proof of work so we have a an option here maybe we want to have CPP and work is the only thing that will find support so even even between healthy and coherent both of us support Authority but different so if you want to do a proper a scientist and that is super important just reflecting to the earlier issue with the code size limit and that's of course discussion I think

we're going to go with that set up for for the Puri wasn't as it is considered to you have this other shape test that maybe even it wasn't then then all of these things have to be with you yeah I got the point my point is that for several transactions who currently transaction fool is limited to so the transactions will already feel turns out that is larger than I don't have any 5k and the reason is that because we kind of store about six thousand transactions multiply that by 25 kilobytes and you have a nice few hundred megabyte memory uses just for a section from itself so

if you if you just do it connects increase on the on the code allowed and we're already in the gigabytes worst case scenario spammers okay we can take this offline I think as far as coming up with like kind of the details of the if we were to do a cross client test set a new one that involves II was them but until then we're gonna have the peer you wasn't test that is there a timeline for that exit might mention earlier or not yeah okay sounds good so next up Alexi

can you talk to us about your idea for changing the network ID and Constantinople in order to make things backwards incompatible yes so it's basically for the screen separation from rosy on classic hard first and in most of the places don't really matter but the one which is doesn't matter is where we're after handshake every point is asking about the doubt artwork blog and the check fix later and that allows you to essentially separate yourself from the waste in classic although they still

using the same network ID the same handshake messages so what I'm suggesting is that to have the change in network ID and actually have a practice of changing network ID every time would be hard work so that we always cleanly separate from the previous network it does have some so we will be able to remove but it does have a need to think about this thing if we do change network ID after hard work then essentially we'll have to disconnect from all the peers and reform the network so that's something to think about yeah yeah but there's another issue let's suppose the time and you know and

I just joined that means that was I'm thinking of I didn't deserve the network IDs and especially if I'd be joining currently I missed I will be on the frontier side of the chain and then I need to already connect to buy something what was that level so yeah I haven't I only just today I realized that there is this complexity about reconnecting we form on the network it would be nice to to do it because you know it's it would kind of make words things simpler I prepare my heart works so currently the

Deaf beauty of protocol supports extending the handshaking with arbitrate field and helix work done that's really hard for a person for something what we can do is that we can expand the handshake so that that includes some arbitrary random data and then peers will know whether they are understanding or not and basically we don't need to change Network IBD just some as somatic material okay I think I'm just gonna need to think about it more but I encourage everybody else to also think about that because I really I think it would really

nice to just make it much simpler there and the attention you have between affords thanks Lexi we're finally to Constantinople so what I want to talk about today was as much as we can get certainly IPS to the accepted state even if they're not they don't have every single thing to find get them to where we're like a green team put them into Constantinople and then start to talk about a timeline so first off let's just go through some of the e IPs and make sure we're kind of all in agreement that something with some of these are definitely going in 145 bitwise shifting we approved of that weeks ago so that

one's definitely going in the IP 210 block hash refactoring is another one that I think we all agreed on is going in the I don't remember if we agreed on to 10:52 the xt code hash or not does anyone remember this one I think this one appeared quite recently so I don't think we did it's one of those which we did not specifically yeah this is next about returning the tech act 256 hash of the contract Oh Lane are you uh we asked a question

no that would not go okay no problem so let's see we've got temple and yeah so we should put that in anyone disagree actually about simple and useful so I if you don't mind I'm going to ask just a simple question about the I know that it's been being posed in the context of the Casper contract right this 210 is completely independent of the Casper contract 210 is located cool so 210 yeah I'm pretty sure we agreed to

tins going in so 1052 straightforward everyone kind of wants it so I think we should do it so we'll say right now it's going in unless someone comes up with a crazy like a reason not to mm-hmm net gas metering 4s store I think everyone's in favor of that Christian yeah I don't think at any point you need to know with I feel maybe

thinking works there I don't think it's simple as the yeah no I was about to say the same thing like it's the mean complexity that I see about this is that you don't have to just meet yet attorney Matt but you also have to maintain a journal for the dirty map and so there's risks involving like what happens if stuff gets reverted what happens if there was like stuff gets reverted to different depths and then rewritten and so forth and like we had to fight with these kinds of issues I see what you know two years ago so my leg ends up I'm reintroducing a lot of trouble if we're not careful the other problem is that this like it does in like it does definitely increase the

complexity of a minimal implementation considerably glad a third problem to the issue it does make me much happier so essentially if we reduce the what this doesn't reduce it again transactions which is completely fine but it means that it will be a lot happier much heavier so I think I posted it somewhere or Anna's asking somebody whether they could put a number maybe we'd have to see that if you were to roll out this Eid Romania now then what could that mean how many processing time so in the

worst case like if we're assuming a daus attacker then this should not make things worse at all because the daus attacker would just make sure that you fill in different fields or make sure not to have a transaction that refills the same field anyway with normal usage yes but I guess it would be good to have more metrics on like what actually is marginally contributing to the Uncle Ray door like margin we could share bidding to load on clients the most right now because it's like to me it seems weird that blocks are being processed in a

couple hundred milliseconds but the local Raiders are is still over 20 percent and that like there are some cases where if if the block its contents turned become closer to being cure simple transactions as we had in December than the local rate what up to forty percent even though the gas consumption was the same so like it would be nice to just have I try to do a bit of more detailed analysis and like see what actually is contributing the most and like maybe it's not even at point level maybe there's something happy weird happening at the animals level and yeah I feel like that sort of stuff will be

necessary if we were to inform any scale in any kind of short term scalability and permits that we wanted to make one testing experiment good to do so one more concept I might have is a propagation is basically a significant sum : and one thing that we try to do is adjust in another four five different machines different parts of the world and just see how fast take at the ball because that might be exhausting being able to see whether they are delays in the what do you mean by virtual machines now basically you should you know I've

machines on the server or I thought about this as well and to make it actually much cheaper to do this experiment I have essentially what it does it does check for intro performance but it does not it relations law essentially don't know the and this kind of stuff thinking about and it starts basically propagating so maybe the internet will be quite tricky

cuz you can reroute you can run it like [Music] yeah I guess bakudo so if you want to use a transformer you don't wanna be with mania history so we can Joseph simply be Sparta no I mean he's down oh there's a still kinda useful to the network because David is somebody still has a distinct complete each oh yeah okay so I automatic experiments okay so

that's given that brought up some questions with EIP 1087 so we'll skip that one for now and not say that's gonna go in or not because we stops a lot of open questions 1014 skinny create - is there anything on that that would be problematic to anybody or stuff that needs to be flushed out I think everyone was in favor of it too cool no comments what but any word about little bit river is just gonna create you ready I think so yeah actually so Nick commented that it needs a

discussion URL but otherwise it's good so there was one and I think my explanation was which might not be right if they create to create to only create the contract if it does not exist arrange so that all of these edge cases out like one or two years ago because I remember we were just like in the GUI theorem code already has a game already has a cost but this particular uses even

though I remember well but there are test cases for the specific use case there are people interpreted I'll gable that's it wrong and I'll get Cod cool so it looks like skinny create two is gonna be going in so that's good so the other thing is do we want to delay the Ice Age whenever we do this for hard fork Ice Age say that one more time now I was wondering if the italic has done the simulation I haven't yet but we can totally redo it although the other thing

actually no we could there's a simpler way to do this the Ice Age was delayed by three million blocks correct correct okay there's all we can do is we can just look at the blog sign chart on ether skin which I'm going into right now and like in terms of block numbers it'll be exactly the same so if we pick up the time at which the ice age kind of starts happening as when the block time reaches say 16 seconds then that would be [Music] let's see it June 5th 2017 and I know that we have about 6,000 like right now

but about 1500 blocks they actually no hold on do you remember what the what was the Dow hard sorry the Byzantium part for blockers at four point three seven million just 127 days you know Kieran so 4.3 4.0 4.1 for 3.9 3.8 so the time we have to worry is when the blocks number starts hitting the 6.7 million range so that's when it'll like make the first spike opposites like maybe less than a second up it's still noticeable so six points it's actually gonna be new months right I'm hold on let's check right now six point seven is still pretty far away all right now we're at five point nine so six point seven is eight hundred

thousand from now and one day is six thousand blocks so a hundred a hundred thousand box is about seventeen days felt like started about six months or so yeah and then after that it'll take another eight months or so until it becomes like really serious so I'm doing it in Constanta and in Constantinople log seems cool okay anyone else have an opinion on delaying it or not questions about the timeline that's we can talk

about that before we decide on the ice age thing so yeah did anyone else have a comment before I move on to the timeline and then we can go back to delaying it okay so we're talking about timeline and then we're going to go back to talking about delaying after so I put a proposed Constantinople timeline that's very very optimistic it's listed on the agenda but basically finalizing the EIP is according to the timeline should happen today the client implementations happen for roughly a month between July 16th and August 13th testing would be August 13th to

September 10th the test net would be September 10th October 1st and then we would launch October 8th that would put us like 20 days before Def Con starts I don't think this is necessarily reasonable but I just wanted to put something down and then we can start estimating how long it'll take for things so Alexi had an alternative timeline that he posted as a comment so I'll go down to that and that puts us having a launch in January most people don't like working during the winter time because they're on like family vacations and stuff so that's why I was willing to get it done before Def Con but I don't know if that's really feasible now depending on which ones go in so my timeline is essentially based

on the fact that you have to produce the conformance test for each EAP before before most of the client starts implementing it so the idea was that we produce one reference implementation for each EAP this one in any client doesn't matter which one then the death directions among implementation allows you to generate you don't really want too many people queue late disasters like so once you got conformance test then you can finalize the specification because the conformance test might actually eliminate some of the problems and so once you've got conformance test and respect then you go ahead yeah maybe I mean I am NOT saying that it has from

I just wanted to be to be kind of work lighting time because at the moment it's kind of smear around different stages right think so implementing this stuff and creating the best for it was kind of concurrent singing which is a nice thing or bad thing is the way it happened well I'm just I'm just consciously you know the the situation changes over time so in the beginning the point implementers probably had more time to do the stuff but nowadays everybody's easy fighting their own fires and they really need to kind of appropriate for the folks back

with very like with a very little chance of it being on implementable before they seriously start doing it and you know I think we should get to that stage so I think that that's similar to how we've done it in the past except we've done things concurrently like Martin mentioned so like we are implementing it in clients while we're creating the test at the same time having C++ be the first implementation that everyone copies from is kind of how I understood it so that's similar to what you put down Alexi and I like that you did it the way you did because my timeline that I put down was kind of waterfall method where it's like

one after the other whereas really it's more agile the way we're doing it where it's kind of things happening concurrently then it could still be the my timelines to be concurrent if you if you look at the this anytime I can go for each et cetera so you don't have to like do it first for the first a theorem for the second we can actually in currently make a reference implementations with all the APIs at once and introduced ones and then you convert somewhere along the way you might not be able you might say that oh you know we're gonna oh go ahead could we set a priority list for those

VIPs and maybe reevaluate according to those timelines so maybe we select five of the easy ideas right now but we say for example bit by shifting is to one block cache is party to export registry etc and the reevaluate as we go which one is from my personal perspective I think advice 50s code hashes uniquely that reveal I don't really see why this would pose a problem not as we think could have suffice a lot has another

family yeah I think this is different than previous hard Forks where we had more like very very high profile high risk changes I think it feels like at least like 210 is kind of entering fundamentally new territory because it's the first thing I knew it like privileged EDM contract that gets called as part of blog processing but even that is like it's a significant change on the surface level it's not like a deep change to how the EDM works hmm I said great however it's not as far

Sun is still fine lines that we D for which 1 to 10 I thought it was final I are not finalized like technically by EIP standards but I thought that the that it was written out completely I might be rather oh yeah okay so the talent just needs to go in and have that then right away another question I had about to but oh we're writing is smart contracts in basic CED my check I know

that I published the I think I published the EVM Anne's the lol for it like I feel like lol was kind of the appropriate level to write it because it doesn't introduce high level jump a very complex compiler risk but it's also readable and you think you publish that somewhere yeah Oh Jack thank you bye bye serpent oh just serpents in bytecode and I could just treat them all for it then yeah if you can do that that would help us I guess understand if it's gonna be a little bit more difficult to implement okay look it's the type of serpent that

seems trivial you can foretell [Music] well what reason I'm asking is because basically we need somehow the audience choose you so kind of water there is one center that will divide so Paul what were you saying yeah serpent I think kind of hard for me to actually move on with this one because every time I propose some change

it's like wait forever I cannot like I'm not able to queue them up easily and I don't want like move forward in all of the issues at one point because one or two feel requests that have to and also we were thinking about using implementation I do some kind of it's something that is quite close to the PMF

codes by the way do we have test cases for this four to ten yeah yeah that's a good mitri question unless someone here knows I remember didn't make but I think in the end essence not to include funny part for the part without words right now which makes it good won't work in high-crime testing because if you

okay so 145 is pretty clear-cut so is 1014 for skinny create two so is 1050 - for EXT code hash I think and then block house refactoring needs some work and we need to discuss 1087 oh you mean the lowering it for up to curve

stuff yeah I think so but like I guess I would like to see like other so first of all is a just guess that could that added optimize the library sported to parity I can't prove though there are libraries as well and like what about other implementations what and you have some solidity benchmarks or sisterhood I can benchmark with it's really trivial to inject them into - just run it against all the clients when we were adding IDIA ec pre compose for Byzantium for the first time we did do the analysis right and we set the forty thousand like and what eighty thousand gasps busted so forth because we wanted

to target like 20 million yes a second right so like there were measurements that we clearly did and you can just read you is up within yeah but what people say is that we doing

I'll lane could you meet a little bit that's marking with my guys okay so that's something you can do once things are more clear you're saying don't have any more clear so for example ZK stock are part of the - trucks I did

I've had those four months ago but those aren't really so those are kind of a complex interactions really hard to measure individual weight of certain emotions okay any other comments so judging by how we have how we haven't really gotten exactly what GIPS are going in yet and the timeline is getting closer and closer to DEFCON it's probably going to be posted EV con before this is released unless we start moving here so if that's

the case we probably should delay the ice age in my opinion because by then we'll start seeing some effects in 2019 from what you were saying right for the ice age okay so and with the earliest we're gonna get done in late 2018 early 2019 this hard fork is what it's looking like does anyone want to push the agenda and make it earlier than Def Con perhaps it's worth thinking about not wanting everything to Constantinople but simply releasing changes as they are ready

let's say if we know that the shifts are pretty straightforward and they pretty much ready why don't you just do hard work for the ships I mean what was the biggest problem with that no that's a key word right so where's it coming from you have to ensure that everybody updated you have to test the whole thing you have to test that exchanges what we want to guarantee a hard work before Def Con then like I'd recommend like probably kicking out to ten just because like we're kicking out to ten and

seeking out 1087 problem probably so you have bitwise shifts and thus can agree to NEX decode hash because those all seemed fairly simple mm-hmm yeah so this can kind of be like a little hard fork if we want yeah what's everyone's thoughts on that does anyone think that it's worth and it's crappy that Nick's not here to defend defend his case for the net gas metering 4s store but I think it was just proposed a little bit too late in the game it's is what it's looking like because there's a lot of open questions about it Wow

to be honest this is why we are usually suggested that everyone who proposes of four or EHD that easy have around since the language because it's easy to argue about so maybe an easy stuff could be a community since we can't exactly it'll intersect and some of the PRI okay yeah that might be a good future future requirement I think the main practical use cases like one enabling white

clients that can sink without literally processing every blog header the other which is like potentially useful right like it would be nice to have a light clients that can sink trust us away in logarithmic time so like if you want to actually achieve that in practice of requires not just the EAP it requires all the other code so my it may well be the case that actually achieving again it's not worth that because we can just rely on the existing approach until he gets like something but through those stakes are like client functionality order I don't know well the other the other alternative would be not to go with a

contract based approach and simply well not an opcode but like I guess a non-color to say yes I just needed well it basically it take the code that was written there in serpent and just implemented in whatever is the code of your whatever the languages that the client was written in no no just simply the whatever your cash now but just do them much to whatever no that's that's dangerous because it makes like it could

make life a lot harder for like clients and it basically means that like the entire historical set of block chains become stuck technically part of the state and like it it nullifies the benefits of kind of automatically having box yet like get hash links the blocks that are much older than they are like yeah like I see the primary benefit of this yep year kind of precisely being that you have a storage tree that contains like some subset of previous block hashes I yeah I really wanted to have to finalize in turn okay in longus and the

spec is considered I just like feel it's [Music] still like way mostly myself finishing this one you're good I would say that I I think I agree with you that 1:45 1052 and 1014 should be priority because they're the seem to be the easiest to implement and what we can do is we can start making tests and start implementing those three and then throughout the next chord F meeting we can decide on the other two and talk

more about them I think that would be valuable and also it would help us to move forward because right now we've kind of been at a standstill for months about what to do with Constantinople so I'd like to move something forward a great person library and there's the option of doing content and Opel soon with these and not changing the the Ice Age and then doing another court earlier year changing yeah that would apply some pressure definitely yeah I think we need a bit more precision in this so first

first firstly the test the state test for ships already and you can already start testing them just have to enabled Constantinople Hartford in facing tests but they already in and second thing is I think Dimitri explained it's it's much easier to decide on the order so we can add another VIP to this to this virtual Constantinople how configuration tests and so we just need to pick one of the two laughs that's

much easier to coordinate this way instead of trying to dump the test forms are you saying create the test it early or make an EIP that defines which ones are going to specify which which which PAP will go next in terms of okay at the same time it would be much much better without decide which one goes I guess like cuz we've had this create to

implementation in some places and maybe we've had some tests for that too so I would keep that like after ships are already there maybe that's like not a full set of test some of them out there I would pick the create two and yeah I think we can focus in a cpp atrium to the final of the implementation of vectors and start to test next week

great yeah I think I agree with that suggestion and basically we have the three easy peons those should be in command and we can also start write a test for it and then maybe one people in command is free and write the tests and see by the report we can actually to make to maybe write a fanciful to debate it for maybe next week next predominately in week to see see what if we played approximately to do that and then based on that we could have a good starting point what do we want to test that sounds good to me anybody have any other comments on this

cool this sounds good we have shifted towards or not shifted towards we're coming together to say that 145 1052 and 1014 are low enough hanging fruit and that had already either been started that we can do and people want a version of 210 that's much simpler and doesn't include like EDM dependencies I just did an extremely quick write-up if you look at number 1218 it's yeah it's like a look at least half of what I think we want but it's much simpler here I'll post it into the little channel can it be turn into mobile without how much can it be turned

into what can it be turned into the bowl lieutenant over time in another court like the fundamental form is kind of the same so it does have a panel some other things in a way that's somewhat different [Music] okay so we can look at that and discuss it for next meeting I'd say sound good sure cool okay so I think we have some good conclusions from this meeting do other people have comments on this and that we don't have a timeline exactly but what we do have is we're starting

stuff and depending on how fast we get stuff done in the next two weeks we can actually start to set some dates and finalize some things like declaration are we a team or four mm oh I think we should aim before DEFCON but that's not final yet because I haven't I I feel like we have not reached consensus on that I think if we if we do reach consensus like if people say we want this done before DEFCON I think we have to only do these three IPs I don't think we can do more than that mm-hmm but that's just my point what is October it's like October 30th to November for

third or something come on we can do it easy for Yankee something I like your enthusiasm and I'll if you're being serious then yeah we can do four okay so let's see if Nick can implement his S store stuff and we can get an idea of any pitfalls that we'd run into Peter do you talk to can you talk to Nick and ask him to do that if not I can mess yeah sure cool thank you explain what we talked about in the core dev call all right anyone have anything else otherwise I think I'm we're about to wrap up all right cool talk to everybody in two

weeks lane or myself will have notes up and I can put in the core dev Channel what the next two dues are bye everybody [Music] [Music]