Conservative Approach to a Radical Roadmap by Sunny Aggarwal (Devcon5)

[Applause] anybody ever thought to call a conservative approach towards a rival roadmap kind of give a certain something like thoughts about people know as well as learnings have you done from Kosmos and how I think but you know things that should be I think should be integrated in you if you point out that could help your reason 20 so you know what about me I work mostly on cosmos and tender man but I also run a podcast called epicenter and run around here called Sitka so just a little bit of background where I'm from so yeah there's your point but the basics many people are probably somewhat familiar but just for a bit more review what we have is we have this multi phase two stuff where we have this like phase I think that space numbers are wrong it actually pays zero but yeah the first phase which is we just basically have this main chain but so the current crew for working through everything that we have today and then we start building this beacon change is provide to this task bar FG state then and we get to phase two then so the speed is zero thing this is what's happening in like a couple of months then eventually we'll get to this point where we start to do shards and we can write data to the charts then the next phase will finally start to execute by the way so far the only thing that has been planned up close now is up to this point this way after this point right the Lotus there's not much you can actually do with these shards yet then finally in the next phase will add the execution engine where we can actually run it um or webassembly yeah or something like that and then finally in phase or we're gonna start working on cross card transactions and this despite this babe has not even been like SPECT or discussed or like yeah so this is heavily you know my main issue here and I think this is going about it in the wrong or I think get into a minute but one thing that's even more worrisome to me is this one way sneaking horse out you know in this phase in the first phase we have this one way burn I have either on the main magnetic area I have to burn my teeth I don't want to use the word peg because any flies being created back somehow I have to burn my ether and then that gets set over and you'll notice it's a different color because you know it's a different type of each like if you can't transfer it back I think that these things are gonna have to difference change values will be treated very differently and you know when you try to move I can't so one way worse this is very worrisome to me so what will it needs wrong here so you know heads up love from the state you know I just read last two and a half years of my life working on for the state but no proof stick has a problems as well yo-yo we don't know exactly how it faces with he said you know in cosmos we see this concern today where there are no I'm partially blamed for a lot of it but you know there are a lot like centralization happening in the balladeer instead we don't know how resistant it is to cartelization we don't know know especially in a roofers fake system where you use the money token as the state how does that affect the economics of the system what kind of censorship attacks are possible one thing I'm personally very worried about is anything probably a group of state can get into later yeah all these things you know I'm very low it's not a mistake but the main fundamental issues of state is it's highly experimental and I just do not feel comfortable shifting a thirty billion dollar network to this one sec untested security model like you know I think that those people eventually we'll find out the answers to these questions but it will take time and experimentation and I don't think trying to jump headfirst because sort of panic moment where it's like oh we feel that we're behind and so we do like just jump into an ex-basketball i think that's making some tasty decisions so I think the problem came when he started this which is the merging of of Casper and Shari I think that you know up until last year up in the last summer the Casper team research team was distinct from this German team and then about last summer or the exact emergencies into you know Jasper 2.0 whatever and so this is where I think the things went wrong and I would say that you know when we're waiting these two things I think starting is way more important than a mistake like you know starting there's an immediate urgent demand need and demand or we need a disability really fruit estate you know it's great but I think it can come in time and so here's my proposal of what I think is a more reasonable way of approaching okay this is what we have today you currently through him maintain we got a proof-of-work and we have an EVM here but first off you know let's focus more on side-chains where you know and that's testy sidechains with Bruce and these proof estate truck so you state your either on the ethereal main chain and then you could become a validator on the side chains and so nobody then move your eats back and forth one of these side chains use them there move them back these sightings that have the EVM on them we already have a lot of people experimenting on things like this day like that POA Network has their X I chain and stuff and so this is really good but they're not using it they're using this prefer Authority I think always you're focused on is experimenting with sort of snake on side chains first also you know a little bit of a plug for cosmo stuff you know if you don't want to use the EVM qualities you can use different types of PM's and systems the furthers of my B's news Christian favorite rates talk on it this morning and really generalize you can allow you to talk with me many different frameworks you could talk DBM change the problem sets a chain of yelling change or parity substrate chain that's really the goal but yeah so let's test proof of State on side chains first then eventually don't then there's already so much great research happening in the process of happen you know I think then the the think working side chains in charge it's a little bit more of a spectrum there's like you know there's a lot of research that's already been going into how to like make more and more trustless side chain and it was really double down on those yes I know I think the scale loves people they've been do a really cool job one come on put some great and here's their you know I don't want I don't want to go too much into their stuff you know you should take a look at how they do it they have this one like one in the speaking manager contract that basically assigns pollinators or different starters on the system and it's running on the etherium for her work shape in the ABM there's also a lot of other cool stuff like you know the map like everyone is working on CK roll-up I just um added as an example but there's a lot of cool people are working on like CK role of techniques the plasma group came up with this OPM system with and some people at consensus and some cool stuff with minimal Viable version senses this guy know but you know having be high low you've been kind of working on a lot of cool ideas and here yes I see your projects I can't go too much into it but now he's working on some cool stuff along the splines job I think we should really be focusing on building these proof of stakes sharp testing who stick at the starters then once a pine past then what we can do is suddenly take off they like you know GBM on there but remove the like plugins where basically stop allowing people to deploy new contracts at the on the route change so now all the contracts that are still there still are running but you know no more deployment of new contracts if you want to complain you contract they go on to the Sharks next after some more time of Douglass's then we can start to allow a system that allows you to easily migrate some of the contracts on off of the EVM onto other channels and then finally when a lot of time goes by like you know maybe five five years or something ten years your first work has proved himself a big women for like ten years I think roofing needs that level of confidence and like history to it before we we should be ready finally we can go ahead and change the prefer work chain to reproduce a chain okay now my proposal for the eighth year whatever of that was a little bit about application specific charge which I think that something which I think it's great that I've seen this actually happening the last couple months now with this whole execution environment idea of that kind of new adopter words it's really good and so I can talk a little bit about why I mean it's great so you know I think smart contracting systems really should be used for contracting not for applications now contracting are things that like are short-term one-time use things need high levels of customizability I SEOs are the like perfect use case for smart contracting because they are short term use things they need high levels of customizability but like you know if you're building index you're building a prediction market system you probably want a more custom build efficient state machine for this you get a lot of benefits when you do this you produce a lot of the attack surface a lot of the issues with a lot of the biggest contract bucks usually have come from weird vulnerability is that like nuances of ppm whether the doub of the pair of Evo's or most of the bugs usually come from weird things with IBM you get a lot more efficiency games because you have a custom build state machine you're not you know you're not this is like you know if you have you didn't have much sometimes this giant opcode system and running on interpreted code you know you have lots more simplistic state machines which makes it much more efficient finally you can find you to optimize patient so you know right now you can't really implement C cash like privacy in the UVM in an efficient way because it's very you know it's not fine tune or a great example accident and so it's missing the cryptography that you need a great example of fine tuning for your application if you want to build a payment system right payment there's a reason Bitcoin was built as a beauty M so system you get those are better at payments because you know they can be parallelized way more easily but no the EBM was optimized for the average news it's not for the specific uses and so for the averages case you need accounts the kids but if you want to build a payment system what we want to optimize your application choose the best design patterns that are very on the patient and so this is why you think you really should be in this dream charting certain claim to not be focused on trying to make all shards equivalent they all run the same Vienna but really we should we focus on applications was a bit starved and this is kind of look like polka dot has been doing for a while that it's actually it's great because I think the theorem is starting to realize this and is shifting towards this with the goal of excuse environment system the other thing is what I like to call the maker dilemma so let's explain why I think maker should be not on ethereal I think maker be on etherium is a parasitic relationship in both directions maker person theorem in a 300-acre here's why so for maker as a die older you are getting the security of the MKR token at the NPR whole is wanted to they could steal all of the collateral from the maker system so you're getting the security of NPR token when you're painting for the security of beef which doesn't make sense you know what painting your terms actually these that eat and like you know you're you're you're the gas bees are like changing relative to the demands etherium rather than the makers of stuff and so you're overpaying for security its corporal in the other direction as well because let's say there's a contentious hard fork in theory on tomorrow that's another pair use accident that runs back or maybe it's Prague powers or something you know who decides which more twins historically I received a TN is probably one of the strongest voices upon which side the floor plans but I say today I think he's probably through the MKR holders because the impair holders can't decide you know what in this contentious pork we're just going to stop running the Oracles on one side of the chain and when they stop circuit stop running the Oracles the dial enough time that just changes crashes and that's in the entire defy ecosystem on that side of the chain crashes and so this is kind of very so this is giving the MKR Dow the maker down way undue amount of influence over a theory of governance and I think this is unhealthy for aetherium in the long term where it gives certain projects or entities such a high amount of influence and so this is why I think by putting the applications want their own isolated charts where they don't have that much control over the governance of the system as a whole I think that can help along with some of these kind of dilemmas and better for they make their system diapers as well okay okay so the next thing I'd like to talk about is television and protocol so the it's a theory on the very much don't they don't they don't want people to not to basically and they don't have this notion application in cosmos what we do is you know validators and delegator in the protocol you can go ahead and delegate your atoms to a valid air and they would basically state for you and then the protocol would keep track but like oh okay these many of my atoms came from this delegator these many of mine came from this delegator whatnot and if this value gets slashed and all the delegator for that provided or go down now what ethereal things that you know we could do this process through smart contracts where you know you could go ahead and you know as a palliative company called syndrome I'm going to deploy a validation contract and people could give each two-month to me in the contract and I validate on their behalf this is suboptimal there's something you see reasons why it's walkable one is you know I don't really want thousands of validators writing their own contracts eventually some of them are gonna have bugs I'd rather have their default one into the protocol but more importantly there's a feature you can only get when you put dummies in the protocol natively so let's see it called in sentry delegation so this is how the delegation would have to work in Nigeria so let's say I have some eat delegated to validate okay and I want to change it to validate repeat what I would have to do is I would have to uninvolved from valid or a wait the entire landmark body period I thought the recent numbers are if it's around three months or six months so I could wait that an Ireland body period then I get the money back in my account and then I can delegate that means I have to go for months without rewards just in order to change my about people who are delegated to this makes validates delegation super-sticky if your delegation in protocol what you can do is just instantly swap dummies there from one founder eight now there be instantaneously and what we can do is we can put because it's all in the same state we can create something called pseudo and bonding period so let's see your delegate the founder a you go ahead and change it to be what happens is you're put into a pseudo unwinding peer evaluator for a while you actually go don't get into being your any the rewards dollar P but let's say some evidence comes that you double sign or something we can slash you because it's all in the same state we can still slash in valgar B even though you're in this to know and bonding period for valid area so this is really important because I think it will reduce the stickiness of delegation you know proof of work centralization isn't great but it's not those it's not that okay because miners are able to jump between my rules as fast as you that's the facile they want to take it they just instantly you know let's say this I'm telling it into a mining with a pool that support this is supporting a bit that I don't like I can this easily switch my hash for a different pool and we need to make it that easy to do or delegation as well otherwise we can occur very sticky validators that that's not going to change over time finally the last thing I'd like to talk about is proportional smashing so forceful slashing is also something that you know I think it hasn't started your dog since we started talking about it but you know the idea is you know validator one that's we should slash dollars more for the larger the volunteer is so let's say $1 I got this validator has 5000 feet and this one has 1000 and they both get slashed this once again slashed at a higher percent 5% rather than 1% so not only do they get more coding slash because they're bigger but they also get a higher percentage of token splash this incentivizes people to not delegate towards larger dollars it sends by to go towards smaller validators but not wait what what why wouldn't I just spend just similar right like instead of running one mile later 5,000 miles for that into my pocket of 1018 so what we can do here to solve this is we also punished more heavily the more number of validators fault at roughly the same time so one validator a 5,000 is double signs maybe we can slash them at 5% but let's say $5 in terms of 1,000 sighs we should slash all of them at 20% so we kind of have to take within an epoch we can say if multiple collar is double-sided in the same key foglar are unlike doing the same he thought we slashed them even more this also has a very nice side benefit that it incentivizes you to D correlate yourself from all the other values so let's say you know many other values are you want to not run on Google cloud because you know that if Google cloud has some issues bill they're gonna go offline you want to be as set like you know separate from that that's walking so you don't get sloshed higher than great and so you know this is the general form of the equation that I think really works it specifies the square of sum of square of roots million formulation also similar to the one that's gonna be like quadratic the raccoon market stuff to quadratic stuff it's very useful formula and so you know we can see it actually gives you a lot of the benefits to one where like you know the more validators are / you know if these are about that we weren't taking into account you know we had 1% and 5% to be okay there's a 15% spot it really impressed me use a formula it actually baptized the / board because two dollars splash at the same time and so you know what we want to do is add some constants to this how do we choose what to have the constants one thing that I think we should do is let's say you are a relative 10% and there's someone else who's a valued at 1% and you go with double sign together you need X classic 29 this is twenty nine percent / but what if the other person is trying to grief you they're willing to lose that one percent but in order to replenish you as much as much as possible so what they could do is they could actually split that one percent into a thousand different validators and get them all the double sign at the as you they lost the same amount but they suddenly backed up by it and these flash to thirteen hundred percent which is crazy and so what we do is this tape should be based off of you should take into the account the inverse Gini coefficient so be more the more unequal that validators the number about is about / in that epoch the more unequal their stakes are the lower that constant should go for the more and the more equal their stakes aren't they should go higher and so that will kind of basically help the fall they thought of the kind of wreaking attacks thank you so much [Applause] you