diff --git a/sig-clients/README.md b/sig-clients/README.md index 8ea9ada1..6a00cbdc 100644 --- a/sig-clients/README.md +++ b/sig-clients/README.md @@ -1,22 +1,34 @@ --- title: "Clients Special Interest Group" +key: "sig-clients" date: 2023-1-09T00:19:20-08:00 --- # Akash Network - Clients Special Interest Group (SIG) - ## Meetings -[Third Wednesday of the Month](https://calendar.google.com/calendar/u/0?cid=Y18yNWU1ZTM3NDhlNGM0YWI3YTU1ZjQxZmJjNWViZWJjYzBhMDNiNDBmYjAyODc4NWYxNDE1OWJmYWViZWExMmUyQGdyb3VwLmNhbGVuZGFyLmdvb2dsZS5jb20) +Meeting happen every [Third Wednesday of the Month](https://calendar.google.com/calendar/u/0?cid=Y18yNWU1ZTM3NDhlNGM0YWI3YTU1ZjQxZmJjNWViZWJjYzBhMDNiNDBmYjAyODc4NWYxNDE1OWJmYWViZWExMmUyQGdyb3VwLmNhbGVuZGFyLmdvb2dsZS5jb20) + +### Meeting #1 + +| Time | Notes | Transcript | Recording +| --- | --- | --- | --- | +| Wednesday, January 18, 2023 10:30 AM PT (Pacific Time) | [Link](meetings/001-2023-01-19.md) | [Link](meetings/001-2023-01-19.md#transcript) | [Link](https://j62h6g4vuygradhil2eeape3a6ojy6vf2ty2orv66m5f6kprsqja.arweave.net/T7R_G5WmDRAM6F6IQDybB5yceqXU8adGvvM6XynxlBI) ## Leadership -- ### Product Lead(s) +### Product Lead(s) + +- Greg Osuri (@gosuri) + +### Tech Lead(s) + +- Joseph Tary (@jtary) -- ### Tech Lead(s) +### Project Manager(s) -- ### Project Manager(s) +- Anil Murthy (@anilmurty) ## Contact @@ -24,5 +36,8 @@ date: 2023-1-09T00:19:20-08:00 ## Sub Projects, Repositories & Relevant Work Groups -The following are projects and work-groups that sig-clients participates in or contributes to (ToDo: Add links when available) +The following are projects and work-groups that sig-clients participates in or contributes to +- [Cloudmos Deploy](cloudmos-deploy) +- [Akash Console](cloudmos-deploy) +- [Client Libraries](client-libraries) \ No newline at end of file diff --git a/sig-clients/meetings/001-2023-01-19.md b/sig-clients/meetings/001-2023-01-19.md new file mode 100644 index 00000000..4a84286f --- /dev/null +++ b/sig-clients/meetings/001-2023-01-19.md @@ -0,0 +1,585 @@ +--- +title: "Client SIG Meeting #1" +key: "sig-clients-meeting-1" +date: 2023-01-18T18:30:00Z +contributors: + - Andrew Gnatyuk + - Boz Menzalji + - Greg Osuri + - Jeet Raut + - Joseph Tary + - Maxime Beauchamp + - Prashant Maurya + - Tyler Wright +--- + +# Akash Network - Clients Special Interest Group (SIG) - Meeting #1 + +## Agenda + +- Discuss the key friction points for acquiring new users to Akash and retaining existing users +- Propose potential solutions to address the friction points + +## Meeting Details + +- Date: Wednesday, January 18, 2023 +- Time: 10:30 AM PT (Pacific Time) +- [Recording](https://j62h6g4vuygradhil2eeape3a6ojy6vf2ty2orv66m5f6kprsqja.arweave.net/T7R_G5WmDRAM6F6IQDybB5yceqXU8adGvvM6XynxlBI) +- [Transcript](#transcript) + +## Participants + +- Andrew Gnatyuk (DeCloud Node Labs) +- Boz Menzalji (Overclock Labs) +- Greg Osuri (Overclock Labs) +- Jeet Raut (Agoric) +- Joseph Tary (Overclock Labs) +- Maxime Beauchamp (Cloudmos) +- Prashant Maurya (Spheron) +- Tyler Wright (Overclock Labs) + +## Notes + +### Key Friction Points + +Friction points that impact Akash adoption and retention are as follows: + +- **No SLAs:** Lack of clear SLA prevents serious players from switching to Akash. Users would like to know what to expect from Akash in terms of uptime, availability, and performance. +- **AKT Volatility:** AKT being volatile is not ideal as a payment mechanism, particularly for long-running workloads as users are either over or underpaying. +- **Unstable Providers:** Provider instability is impeding retention as users lose trust in Akash. Even though users chose the provider and are expected to research the provider's performance (uptime, deployment count etc). A general lack of a good way to choose the best provider often results in user errors. +- **Lack of Private Registry Support:** This prevents an enormous number of workloads from using Akash. +- **Lack of Automation:** This makes it hard to keep up with critical changes such as provider drop-offs that impact a deployment. Spheron for example has to update the ELB configuration. +- **Lack of Storage Persistence across Leases:** The lifecycle of Persistent Storage is limited to the lifetime of a lease making certain data-intensive workloads extremely hard to debug on Akash. Resyncing of blocks from scratch for moderately heavy blockchain node takes anywhere from 20-60m, for which the deployer has no option to wait. Unlike in the classic cloud where one can attach a storage resource across deployments, so even if one deployment fails the workload doesn't have to wait to resync when the drive is reused. + +### Potential Solutions + +- Standardized Provider Attribute Schema: This will allow users to compare providers based on a set of standardized attributes. This will also allow providers to self-report their performance and uptime which can be easily rendered to the client. +- Using Smart Contracts for Automation: Using a smart contract network such an Agoric can give automation capabilities for Akash. This will allow Akash to be more resilient to provider drop-offs and other critical changes. + +### Updates + +* According to Joe Tary, the Akash console will be ready to be opensource by end of day. The source code will be available to the public on GitHub after stakeholder review. + +## Transcript + +_This transcript was computer generated and might contain errors. People can also change the text after it was created._ + +Tyler Wright: Absolutely. Cool. We started the recording process, and we've also started the transcription process. So again. Hello, welcome all to the first SID clients meeting. There might be some folks that are going to be joining us late, but we're just going to get this thing started just for the second time. Again, for the first meeting, the structure will be a little bit looser than it will be as we continue to move forward as ideas and projects continue to form added this sick. Then it'll be very clear on how we want to handle these monthly meetings and where we want to spend time. + +Tyler Wright: But for the second today, it's really about folks. Getting to know one another trying to understand what the sick is about. Obviously sick clients. So whether it be any of the deployment deployment tools, or any way that folks are building extraction layers on top of a cost to do deployments. We would like those kinds of minds to be participating in this state to be talking about things as to maybe what they're working on their specific project. Some issues that they're running across, you know, again what's on the horizon, so that this group can continue to work together to push forward, client, client libraries and all things clients on Akash. Anyone have any? Thoughts or questions about that? + +Greg Osuri: So to start off, right? Like + +Greg Osuri: I think it's fair to start with. You know what is number one preventing people from using Akash? From a client experience. number two, was preventing people from retaining. + +Greg Osuri: I think these are two set of broad sort of like areas we can start off with. and, I can actually open up the floor Maxime here, creator of Cloudmos. How are you measuring? What kind of feedback you're getting? and, Can you enlighten us a little bit Max? + +Maxime Beauchamp: Yeah, sure I can give you my thoughts on these two topics. One is a little bit of a little bit hard for us to measure, like why? Aren't people using a cash? You think a big topic? Big question that Like we we measure a bit, the usage of the users and we get feedback from from people using combos. And we also use it. So we know a little bit like the pain points using a cache and I think you guys probably know it as well, maybe it's the + +Maxime Beauchamp: not having like an SLA or like making sure the your deployment or whatever you're running is always a up and running like we run a ton of stuff on a cash and sometimes for whatever reason, the provider closes the lease might be a value. They're not a valid reason but it just happens and we + +Maxime Beauchamp: Over time kind of build like a written. I don't know how to say in English, but just a bit more friction. In the way we use a cash in general, that we again, we know we can't run this because we might wake up the next morning in our thing, not running. So we just run things that we know and we accept that at some point. It might not be running anymore and I think it's the main pain point for keeping people on their cash but for attracting new users I think it's a whole another topic. Yeah it's hard to say Well there's a ton of stuff ton of fiction, many around crypto and I don't know about yeah, attracting new developers that are not Web3. + +Joseph Tary: You. + +Maxime Beauchamp: Knowledgeable stuff like that. And that's a big fiction. But otherwise Australia, other problems that I'm not thinking about right now. + +Tyler Wright: Go ahead, boss. + +Boz Menzalji: Yeah. Question Greg, I just got off the call with Sharon Prashant, son, and this is exactly what we're talking about for the last half hour. So Prashant, if you want to share your experiences of what you think would is preventing you from, basically onboarding, the demand that you're seeing on to a concert, that would be fantastic. + +00:05:00 + +Prashant Maurya: yeah, racist to me if my voice is horrible because I'm kind of sitting up + +Greg Osuri: Good. yeah, you + +Prashant Maurya: Hello. Yeah. + +Tyler Wright: You're on mute now but we could hear you prior. + +Tyler Wright: Percent, if you can hear us, you're unmute. + +Prashant Maurya: Hello. + +Tyler Wright: Perfect. I can hear. We can hear, you know, + +Greg Osuri: Well. + +Tyler Wright: We can see you now. + +Prashant Maurya: Okay, so let me just make me all soluble. So just to repeat again from starting. So, From our business. Point of view, we started targeting two markets. The one market was around like low cost RPC market where people like, in a developing state, they wanted a low cost RPCS to run. We went to hired with that market. We looked into that, we tried to propose a solutions and then in return of that, we were part, we now partner with suite network, we got partner with SCA Network Blockchain, we got partner with Kaive, which is going to be announced Soon. All of these things have happened but what we have been seeing so far and then there was a Actually we'd weren't talking to Reef if you know, Reef change. So, Reef was running there. One of the RPC workload on us. Somehow one of the providers just got shut off, because of that. He had to again, releasing the entire Blockchain and + +Prashant Maurya: That was like pain for him. He said like, guys, we can we can deal with Like, we can deal with the speed. We can deal with all of those things, but we cannot deal with that taking 10 minutes or 20 minutes, 30 minutes just to leasing the entire block. Which can be easily solved solved by a persistent storage and that is where we are also, like trying to figure it out at Firon as well, like, how we can make the rethinking also much faster. So, for example, either you just do a persistent straight directly on the same machine or we can connect the mount, the other machine, which is, which is somewhere else with the faster input. It's it's again, open for discussion like how you guys want to take it forward. The second major problem which we started seeing was around. So we were in a talk with some services companies because as I said, like the market was around like, where people want a low cost compute, + +Prashant Maurya: And we went had to service companies and then we had a discussion with them with them. Also the same concern the for them, they load was not a concern but the concern for them was around the private deployment part. So most of the service crews companies work with the clients and the clients are they have to adhere to certain laws and clauses which is basically like they cannot open source their products until they get the permissions from the client. and, Currently akash don't have a public docker like private docker support or any kind of thing. So that something just creates one more bottleneck for us to scale vertically? Okay, the third and the most important part was around the stability of the node itself. There were a couple of note providers, who just randomly shut down our workloads are + +Prashant Maurya: We were not knowing like, what exactly the reason was, we are just trying to what we have already planned to schedule a new call with the Mars and a few more tech team to try to deep dive into why they shut it off. But there was unstability which we have seen. There were a couple of providers which are working perfectly smooth, but the moment you increase the workload. For example, let's say you want to go to a terabyte of SSD are like hard disk or storage are So if somebody picks up then possibility of shutting them, now is kind of more and we are still figuring out like, what is their configuration, which where we are seeing a drop off. So these are the major areas where we are kind of seeing the difficulties to scale. And from our end like we have done our best to make sure like people are getting redeployment features with an early getting the auto auto like auto really deployment, people are getting so we are able to solve those problems but again the + +Prashant Maurya: Once you work with load, balancer, let's say if you are working with the load balancer, right? Load balance, need to be getting injected with the specific IP, addresses, or locations, to basically load the balance. It's out. But that becomes very difficult for us to handle that as well. So, how we can either, we have to just take the entire control of that and then pass it on every time which is not even possible. Like it is, it is, it is your hip hop to solve that. So, these are the couple of pieces, which we have phrase, we'd love to like, wanted to just put it forward as and I've already discussed with Boz as well. So, Yeah. That is it from our side? + +Greg Osuri: Bouncer, I don't think I understood that. + +Prashant Maurya: okay, so let's say Generally what happens. Like, for example, I just firon are most of the load are most of the loads are learning behind the balancer, right? Which has a geolocation based distribution system where it redirects the load. Now, let's say if g location one, just got shut off. We have to every time that load and if it is a AWS system, you don't have to do anything because it is like the moment you create, you just map it with the route 56 and just take care of the rest. But here if you have to do it manually then we have to basically create one system here and then every time you have to change the provider. So for example, if that provider shut is off, search it off and then we rebued it. Most likely that same provided, not pick it up, right? And then port might also become different And that mapping kind of creates a problem again. So, + +00:10:00 + +Greg Osuri: Sure. + +Greg Osuri: You got up there. You know about starting with + +Prashant Maurya: Right? So every time again, it's a manual work right. Which which causing which, which caused the problem again. So that is weird. That issue comes into the picture. + +Greg Osuri: Okay to do it. But I mean I understand where you coming from,… + +Prashant Maurya: Yeah. + +Greg Osuri: right? It's Well, Amazon doesn't do it automatically. Maybe they're like scripts now that does it. But if you're running EC2 instance, for example, on Amazon, right? You have to monitor cloud, watch and your cloud watch will do the auto scaling groups, which in turn does the configuration for your tool balancer. So it's still I guess it's sort of a click, click click on automation, but like underneath the hood, that's what's happening. + +Greg Osuri: And where are you? Running your load balancers. Like you're asking + +Prashant Maurya: Sorry. So as of now,… + +Greg Osuri: Where are your load balancers running? + +Prashant Maurya: it is for sure on audible is no other option. So that is where we are. We are setting at + +Greg Osuri: You don't run your own load balancers. + +Prashant Maurya: No, we are using the exactly as of now. + +Greg Osuri: Cool. Thank you. + +Greg Osuri: These are very, very I mean this meeting is not about solving problems, it's about just sort of like gathering problems, right? Quite a lot of these. I mean, Recenting blocks is something that has hard, but I think like, quite a lot of the problems that mentioned like load balancer thinking can be assault through automation on a client itself like using a managed service but the question is like how many of them know how to do that? And number two, how easy is, It's not a trivial thing, right? It's definitely. It's quite a lot of automation in place. UM, and the lack of events or events on Akash are very rudimentary, very early stages. So + +Greg Osuri: So the only option is really polling for deployment and hopefully you're pulling interval, is real-time enough, that doesn't make the user, you know. Processor distraction in load but ultimately I think it's sort of like complicated eventually so making that process easier is beneficial for for everybody ecosystem. Cool, Joe. So + +Joseph Tary: Mmm. I mean I think everything that we know as far as like you know, what's up with the the console it's kind of the same point. Everyone else that I'm talking about Consoles, guided its own little, you know, keeping problems because it's relatively new code base. But the You know, General, you know, since they're about especially private repos, I think that's probably going to be one of the biggest pain points for anybody who wants to use the cash at scale. Um, you know, it's great if you have an open source product, and you're relatively comfortable, putting your repo, or your container, at least out there publicly. But you know, a lot of people are gonna least want to in early stages of the thing. Keep everything kind of close to their chest, to be able to use private docker container or docker repos. Some sort. + +Joseph Tary: So, you know, having a good facility for that and you know, especially that's where the clients are gonna have to step in and be very handholdy with people because setting that stuff up in Yaml is gonna be very intimidating for people. So the more we can do on the client side and also even for something like the automation stuff, Like one thing we kind of probably want to try to avoid is have all the clients have their own separate ways of dealing with things like You know polling and automatically deployments and stuff like that because we have people who are we come to the benefit of being distributed blockchain If you're you know redundancy and robustness system is not on portable between different clients. + +Greg Osuri: Hmm, good point. to summarize you, you're a sense you're saying, like + +Greg Osuri: If you have, + +Greg Osuri: Portable. So Akashi, you know, really what we're the vision is to create a super cloud in the sense that doesn't really matter where your resources come from, you can have a very hydrogenous or like, Like resource like, allocation or acquisition. But you get a homogeneous layer on top of that, that lets you manage a cloud. Regardless of where the supply comes from. So, in that way, like yes, pushing automation to the client versus pushing automation into a protocol level in terms of, you know, load balancing or not. Makes a protocol like if you push the automation on the protocol level, obviously makes a protocol a lot more portable regardless of clients. But if you Like our expecting a clients to do their own implementation falling or watching or whatever. Each of them are going to have their own sort of like, + +00:15:00 + +Greg Osuri: You know, approach and then as a user that user is stuck or like locked into the to the client's approach. Right. I mean I guess the question here is like… + +Joseph Tary: Yeah. + +Greg Osuri: how much of power the clients should have in terms of like critical elements that ensure the stability workload versus protocol itself, right? + +Joseph Tary: Yeah, that's that's kind of what I was getting at is just, you know, Oh, yes or two. I think there's a push back on that and sense that we also don't want a busy up the the protocol with a bunch of stuff that is really very much specific to especially like one specific client and it's approach for things. Like We don't want to run into a situation that you see with some protocols where they're just absolutely, you know, disastrously complex because of all the accounting for very unique use cases. + +Greg Osuri: Correct. Correct. I mean that's where you draw the line, right? And that's the hard line to draw. It's not like we can't build, or we cannot build. um, it's like big government versus Small governments, right? I mean Broadway to say it… + +Joseph Tary: Yeah. + +Greg Osuri: but I think like Or it can make the same argument, like, how much power does a states have versus how much power does? The federal has right? Like + +Joseph Tary: Yeah, I think there's also may be some things we didn't, we can do kind as part of this working group to to find a bit of a compromise between it where instead of necessarily pushing it onto the protocol layer for a cash generally speaking we could have something more like a standardized way of defining some metadata on top of deployment saying sort of like how many replications across how many providers you want or what kind of like status monitoring and redeployment things you want. And it's not enforced by the RPC nodes they just transmitted as metadata but the clients all kind of agree on how that data is defined and support. It sort of just like an open standard. + +Greg Osuri: And sort of like how, how the What kind of attributes in terms of what kind of schema you know your your agreeing to? I don't know… + +Joseph Tary: Yeah. + +Greg Osuri: how you consume your performance data of some sorts. + +Joseph Tary: Yeah, that's really that like, you know, you say, you have just a field, you can put in the SDL, which is sort of like a replicas. But yours like specifically, say, I want to replicas of this deployment across two different providers. And then the different clients, may implement their exact, like, how they do that slightly differently. But they all understand that means that you want this deployment to actually exist as two separate deployments on two different providers and they kind of all manage it, you know, their own implementation of it, but they understand what the the definition of the requirement is. So if you go and take that same SDL and you want to replay using a different tool, again, you may have to like, you know, go through a slightly different setup process, but the end result is going to be the same. You don't have to go through and read design your entire redundancy strategy inside of a different tool. You know. + +Greg Osuri: Right. Right. Cool. Let me yeah, that's good way to put it like + +Greg Osuri: also like bottom line, I think the whatever automation we do the challenge, The aspect shadow new aspect is, is security in portability, right? Like Like right now, I + +Greg Osuri: You know, say I want my workload to be load balance, I got to trust the client with my keys. There's really no way to do it,… + +Joseph Tary: Yeah. + +Greg Osuri: you know. So that fundamentally breaks the the security nature of cash blockchain, right? So,… + +Joseph Tary: Yeah. + +Greg Osuri: you know, we can to deploy. Yeah, you can deploy, but in order to actually deploy and reliable way, you got to custody or tokens or custody of your wallet, which is not great. um, you know + +Joseph Tary: Yeah, that's a problem. I think all the clients we need to have a fairly good way of figuring out how to do some form of, you know, delegated tokens sort of thing, where you can give some some facilities to the client. But you know some what restricted Use case you have to give them your full like you know like root key they can have like either like a derived key or a certificate or something like that. They work with that can be revoked in the event that you know longer trust that client and it doesn't give them full access to everything. + +Greg Osuri: Right. Turn over. + +Maxime Beauchamp: maybe can be used with, at the + +Greg Osuri: correct, expanding on, on a Essentially what we want is a revocable. Key. + +Maxime Beauchamp: Which is what are the animals you to do? but I think that We talked about it with Adam,… + +00:20:00 + +Greg Osuri: The granular. + +Maxime Beauchamp: in the medicines back in September. The authentication with the certificate that can be improved. and he had an idea to change it entirely with the like, Getting rid of the TLS SSL certificate and using the Blockchain instead. I can authorization or a signature with the Blockchain. I don't know exactly what he had in mind. + +Joseph Tary: And I think you used to keep using jaw authentication. So you'd like basically set up actually took some prime minister remembering it. But yeah, I remember Job we There was some discussion of replacing the TL certification with John. Which would be easier and menace to the Blockchain. + +Maxime Beauchamp: Yeah. + +Joseph Tary: Yeah, the TLS is problematic for clients, especially if you have a web only version of it. + +Maxime Beauchamp: Yep. Yep. + +Tyler Wright: I just want to add, I think it's great discussion. I know some other people probably have to speak from a fundamental standpoint. This sink is designed to, not only have to kind of discussions that we're having, but if it needs to turn into a working group, for more discussion to be turned into a PRD. This is a represent, some of the ideas that you're talking about. So whether it be private containers, you know, this thing is supposed to think about what product containers to be talk about like how that problem could be solved and then implemented solution. Be at this thing. What we can do to improve the network. If again,… + +Joseph Tary: Yeah. + +Tyler Wright: we need to collect the working group to figure that out and that's great. But I just want to make it clear that these are great discussions that we're having. But these next steps that will have between meetings or going to the next meeting, should ideally be prioritizing the problems that are facing clients. And then working as a larger group to solve those problems with code, + +Joseph Tary: Advocate just for my classification. Is that the purpose of this? We're working groups and we're talking about clients, are we talking about clients as in people or telling clients as in software? Okay. + +Tyler Wright: Clients is in software and this is a special interest group. I just want to be + +Tyler Wright: listening that we might have a specific objective, they're trying to solve, and and the deliverable that a working group will put out Is this PRD. Whereas, this is a special interest group which has like again, console Cardinals sphere on other clients that will be added will also have a voice here and any other community participant. And this SIG is directly going to be implementing solutions that are going to further the Akash network. I know that Overclock has been thinking about private containers for some time. I know that next thing just talked about it Prashant just talked about it. So as the next order of action, whether it be a working group or whether it be a specific project, we should be talking about next steps to actually, Talk about making private containers a reality. + +Joseph Tary: Yeah. I think that does need to be probably a little bit bigger than the scope of this special interest group just because private containers is hardly something that's client specific. That's a general Like at the very least, we need like Urter atom here hits. It have any sort of, you know, chance of making good movement on that. + +Tyler Wright: Absolutely. And so this this clients might say, Hey private containers is really important. + +Joseph Tary: You. + +Tyler Wright: US pushing client forward. So we need to make that a working group and Adam and Artur from or other people from other states like to turn it like Yeah, private containers is extremely important to us as well working group by the developers. + +Greg Osuri: Here. Before we jump on defining, the roadmap I want opportunity to To get,… + +Tyler Wright: Mmm. + +Greg Osuri: everybody's input. So, this SIG is not about this. Signaling is not about + +Greg Osuri: You know. Determining a roadmap and saying that we got to go work on private containers, right? + +Tyler Wright: Yeah, I'm just using that as an example. + +Joseph Tary: Yeah. + +Tyler Wright: I'm just talking about + +Tyler Wright: Yes. + +Greg Osuri: Yeah. So what about just payments and settlements, right? So like stable settlements has been quite a lot of like demand from us, but Maxime I want To understand. How you're measuring like your conversion rates. Like, do you see? People dropping off when it comes to like, filling your wallet, right? So the download your client And then they have to set up a wallet and you know, get paid to deploy. What is a drop rate? You're seeing + +Maxime Beauchamp: to be honest with you, we do measure some stuff, but I didn't went to the extent of like Having a funnel and like measuring each step where users drop off. So I wouldn't be able to. I I think we have the the data to get this information but I don't have it right now. + +00:25:00 + +Greg Osuri: You think you'll be able to get us some of the data for the next thing. It's one understand like You know how we like? + +Greg Osuri: Say it's what I understand. Like, How are we prioritizing features? Right? Like so, we have like tons of things. We need to do on the same, on a client side. And I want to establish a framework for prioritizing these features, right? The reason for that, like, What do we choose? Like, Do we go implement private containers? First? Or do we implement like, Stable settlements. First, you know, + +Greg Osuri: So the more data that we have a better. So now, of course, with console, we're hooking up the, the floor as well, right? So, I understand once we launch console, what the drop off rate for people are when it comes to like, you know, Getting tokens to to work on deployment. and introduce things like, you know, things like a faucet program that will automatically feel give you some Tokens based on Odd Z,… + +Joseph Tary: and, + +Greg Osuri: right? If you don't deploy them within like I don't know, a 30 days, you ought to see Grant is is evoked or some sorts. So, like a lot of things that we can do. + +Maxime Beauchamp: You can put a date in the + +Greg Osuri: correct to encourage people to like deploy, but I do want to understand how much of a friction point is not having stable settlements actually resulting in. + +Maxime Beauchamp: One way that could be done, I guess is to have a public roadmap that people can vote on a very certain features. And that they would like to have for, I guess. + +Greg Osuri: Good idea. Roadmap. Public voting. + +Joseph Tary: Yeah. + +Joseph Tary: can also be useful for the client developers that we know what are the the points we're trying to get metrics on like if we'd be suspect that Stable settlements is an issue. You know, that gives people who have clients, you know, some direction as far as where they want to start, you… + +Greg Osuri: and, + +Joseph Tary: querying their users and trying to find out where their pain points are rather just saying, You know, what's your problem with it is like Do you feel intimidated when you get to this part of the deployment and you know it's more pointy questions to you like better answers are customers. + +Greg Osuri: yeah, and also got to be careful and like how these boards are getting like you know, Like who's working on these things, right? Like if you go as like talking community,… + +Joseph Tary: Yeah. + +Greg Osuri: what do you want stable? Settlements are private containers? There will never understand what private containers are. So it's like, I'm educated. + +Maxime Beauchamp: simple form to fill in for like, A bit of a bit of information on the user that is voting. To be able to vote. To be, to be to have like, quality votes, and not just like, people randomly voting features. I don't know. I think it's a better way than, like, trying to guess. What people are actually looking for. + +Greg Osuri: Absolutely. It's one way, right? We got a,… + +Maxime Beauchamp: Yeah. + +Greg Osuri: we gotta think about like, like that. We have the energy area water problem, right? Like that's why we have delegates delegates you know, waters elect delegates and delegates to the voting on behalf and delegates are supposed to be the gatekeepers. Of such critical things, right? That's why I believe in like delegated democracy as well. I don't know if you into that stuff, but it's pretty interesting stuff, where you? Yeah. We can elect experts but then question becomes like who who are the ones? Who are like what experts themselves right? Anyway + +Greg Osuri: And I don't see clients is, we're touching on a lot of things liquid democracy. There you go Jeet. We touching on quite a few of Like elements because SIG clients is where the user begins. So you talk about private containers. I know that goes across the scope of this, this, this the SIG but at least we should start. um, you know Renting out and listening everything on the table first. And then, you know, depending on that we can we can Can have other things take over from there, and, and prior it as accordingly, but hey, thanks Max. + +Joseph Tary: Thanks. + +Maxime Beauchamp: You're welcome. Thanks, guys. See you later. + +Joseph Tary: See you later. + +Greg Osuri: There. What about? I don't know how many of you are here with Dows and whatnot. We love to have like Jake Hartnell. And those guys come in and give some input because they had some requests about um, + +00:30:00 + +Greg Osuri: internation accounts where you have essentially a account on different blockchain on a deployment Akash, So that gives a very native experience for the consumer chain to have like this, like, white label deployments, right? So, use cases that are the proposed was, like, Look Juno users do not Dow users. For example, now, being a smart contract which essentially is controlled by by talking itself has an account, right? A Dow can control the the deployment directly on a cache. That way you're guaranteed, there is no malice in the supply chain for deployment. so it's very incredible because today a big challenge where defy on general with Web 3 is DNS is attack. Vector Prashant. Get it. So question. + +Prashant Maurya: Yeah, I I don't have a question but what you're pointing it out? Now we have already been started working on it for multiple change for UID deployment as well as for compute deployment as well. So most likely within that is what we have been also working on, is to how to enable the Dow based deployment and control model. We're downs. These compute rather than individual or owning it. So we have been researching about it. Our contract has been audited and It's under process. So, just,… + +Greg Osuri: Oh, it's actually a contract on… + +Greg Osuri: which chain. + +Prashant Maurya: So we started,… + +Prashant Maurya: we are starting with polygon we are going on. So once it's done a polygon, which means like, as usual we scale work, like, originally on all the other chains of AVM compatible, And then it's up to like then on you evm chains. But as of now even + +Greg Osuri: Alright. So so to to repeat, so an account on polygon can actually control a deployment say,… + +Prashant Maurya: Yeah. + +Greg Osuri: on Akash and your own polygon itself using a smart contract. It's incredible. + +Prashant Maurya: Yeah. + +Greg Osuri: It's awesome. There's a exactly what I'm saying. So to summarize what the problems that in general, right. So today a big attack vector is DNS. You know, so whoever you know, controls a DNS, they are. They're at, you know, that's the biggest attack point, right? So if anybody wants to censor or attack, A crypto project, they go and attack the DNS through social engineering or through regulatory enforcement or through whatever you want to you know, point out right. So that makes it extremely sort of like fragile because nobody actually accesses You know applications using the smart contract directly, they actually use a web-based tool, right? A client + +Greg Osuri: So removing their attack work remains. You have to shift the the trust domain away from DNS to something that is acceptable like on chain, right? So in that brain we were like like experimenting. How that would look like and one one area would be like if we can Move the entire supply chain on a chain. So what that means is all the way from code production, right? Assume, like Something like, Gitopia, Right Gutopia where you have, you know, which is like a github alternative, where you have cold contributors contribute and that could call an action using antigen accounts on a cash network to build the workload, the application, you know, publish that assets in a publicly very viable, you know, storage network like RV. + +Greg Osuri: Those assets can be then consumed by and process using Akash network or any other. You know, network that can actually process a short asset and the entire, you know, flow is completely recorded and audited on a blockchain, right, or across different. Blockchains of a matter of fact, So using a browser plugin, we can sort of give that guarantees in terms of like integrity of the supply chain. So those one of the ideas that I had or like, you know, you're like brainstorming and the GPU. That's how you shake your head. We've got thoughts on that, so you're muted. + +Jeet Raut: Oh sorry, I mean I some general thoughts about this but please go ahead. First. + +Greg Osuri: Yeah, that was just my my, you know, flow. Prashant that before you leave your very thoughts or like are you thinking about like supply chain or you thinking more mostly like just on the, on the Internet aspects of it. + +00:35:00 + +Prashant Maurya: I'm not on the on the supply chain here. So for for us it is more around the usability part which will keep on focusing. + +Prashant Maurya: Happy to add more value. I have added Greg. You you are not on our personal group. I have added updated few numbers there. Before I drop up, just to give you guys some numbers, which we have seen. Like you were talking about now. + +Greg Osuri: Mmm. Okay. + +Prashant Maurya: So we had done around 600 42, total attempts of deployment on the container side without any marketing. It is a closed or thing which happened with the users out of that, 300 got closed. Meaning they were activates at point certain point of time. Like it can be close by user. It can be close by the provider. We have to just give you the exact numbers 370 field. To start because of it was our fault and it was bug or some system issues. And 25 is currently heavily workload, which is kind been active now. + +Prashant Maurya: And these are being done with with like in in a very close loop. This went on audience so far, we have not. + +Greg Osuri: Got it. + +Prashant Maurya: Yeah, so yeah. So these are numbers. I have already put it on discuss so you guys can also take a look. So yeah, thank you so much guys. Good to me. It's already it's it's late at night here. + +Greg Osuri: Thanks. + +Tyler Wright: Yeah, thank you. + +Greg Osuri: Great. Good. So like the Dow own deployments of our data services. For Dallas is another area. + +Greg Osuri: And another area wanted to explore is. + +Greg Osuri: I think interchange accounts also opens up like opportunities for other blockchains to have composability overakash. + +Jeet Raut: and, + +Greg Osuri: I like Agora kid now directly. Deploy and Agoric, could be like the coordination chain, but across all chains, where you logic is written in a small contract. And, and each of these resource change like a caution, others are just providing resource where you coordination happens in a guarantee manner. I'll go. Right? That's another and written in JavaScript. So it's very portable and it's easy for people to understand. Um but I don't know area with understand accounts we enabled in it for but that had its own fair share of problems. So we had a reward between advice. So this PTSD in our teams about international accounts which for me to make a solid case I need to + +Greg Osuri: Present present a lot. A lot more justification. + +Jeet Raut: Yeah to quickly go out that. I guess that's kind of my question, right? The two ways I can see Agora really working with the cash on this. I mean primarily on this versus ICA we spectat about me for auto read employment. It's been almost like a bit of a year ago. Honest we had a developer build that out it works but obviously I guess that's back in the day I don't know if it's functional now. + +Greg Osuri: Oh wow. + +Jeet Raut: With ICA. So, all that to say is I think for us If there's any and even think about the doubt question, right? That's something like I want to. We want to build out really robust. Now doing we think that's a like long-term that'll be very useful. even on our roadmap, you know, like I manager about Agoric, There's a good use cases like this. Like we're happy and it's something we can work with you guys on as well. We could like Bonnie cost, There's things that are be useful for users and if it's the work is a way to do it. As you mentioned, as a coordination change. That's something that we can definitely work on. + +Greg Osuri: Yeah, yeah. So they're little limitations to toward a proposed in the workflow. The big challenge is there is an off chain component. While pushing assets, right? Because, you know, the asks of private right in terms of, you know, the actual source code, the action manifest file. You know what, what workloads you're what containers, you're running all that is private information, so that happens, you know, directly in a period of your manner between the client and the and the, and this and and the provider. automating that this in order the two ways to solve it, the first way is obviously, if you can Enable some sort of TCP interactivity using using a small contract. Sure breaks determinism but you know that part of the workflow, it doesn't need to be terministic. Right. + +Greg Osuri: So we can explore that one area. How do you? I think that's general problem. Like, I think like Ethereum has this or really has, like some off chain, like invocation, Tcbn education, right? That's a capability. I think core we need to and smart contracts. I don't have a few. If you work is working on it, I know causing awesome had some approach. They're thinking about like, how to do like, You know, off chance services service access like that's needed for oracles as well, right? second is, Just it's second is just a new type of deployment call on chain. Deployments where he if you could make an argument saying that well if you're a D5 project if you're if you're a web you know a web layer for a defect project, you're a fully public project, there's nothing about you, that should be private. + +00:40:00 + +Greg Osuri: So, it makes sense for you to actually build your entire workflow or your log files, all that to be public. So that we have absolute guarantees that you're not logging what you're not supposed to be logging and you are, and you build process is not compromised, it's fully auditable by a public, right? So that has it's it's benefits as a public utility. So, you know, those are those are that's one big challenge that that needs to be solve for full automation on chain. And and We have to approaches and we've got to decide on one approach, maybe both are beneficial, not sure yet, but Yeah, that's what we are. + +Greg Osuri: Cool. Joe, you want to give us an update on console where things are + +Joseph Tary: Yeah, it should be the general goal. Like, I think we're looking at open sourcing it probably within the next 24 hours, so We'll be ready to open source. + +Greg Osuri: Oh, wow. Okay. + +Joseph Tary: So you know, we've been wanting to do this for you know last couple of weeks and the major hold up is just there are certain parts that we want to make sure in a good state if other people are going to base code off of them. So Yeah. + +Greg Osuri: Of course. + +Joseph Tary: The last little bits of that are just occurring over. So hoping to actually have done yesterday, but there was a few parts in the STL spec that I have to improve kids are failing for an STL, so I don't want that that code is the one. I'm most interested to make sure it works well because I think a lot of people might just want to grab that instead of writing their own STL to manifest code. Um, but hopefully that'll be all done today and… + +Greg Osuri: What? oh,… + +Joseph Tary: then we can open source it later today. + +Greg Osuri: if we're that close, I want to Take a stab at the Readme file. At least. I'm really gonna read me files. Haven't read me files. + +Joseph Tary: Okay. + +Greg Osuri: got my report. Trending on Github Golang like three times. + +Joseph Tary: Okay, let me the code will be ready to open source if we want to do actually. Yeah, that's good point because we need to contributing dot MD before we open source and so yeah. + +Greg Osuri: License and all that other easy stuff. Let me handle you, do the hard stuff. That's cool. + +Joseph Tary: Yeah. + +Greg Osuri: That's exciting. So again this is amazing that we're actually open sourcing first before launching the product. Which are, which is, which is amazing because now we'll have the world. Take a look at an audio code before we say,… + +Joseph Tary: Yeah, find it. + +Greg Osuri: it's all good. + +Joseph Tary: Find all the things are concerned about for the we launch it and they have concerns. And there's some stuff that I think, you know, from the discussion today that I think, you know, maybe for the next meeting is a good thing to to sort of put on the agenda. It's just that there's for all anybody who's writing a client for a cash. There's probably like a certain list of like, honestly toward activities or bits of code that everyone's gonna have to commit to write that is fairly specific to Akash. And if we can isolate those and have just good open source, implementations in a variety of languages, it'll probably expedite other people writing clients and integrations with different tools. + +Greg Osuri: Yeah, what happened? Luna was working on a project. Why is he not in the SIG? Or do we invite him, or we forgot? + +Tyler Wright: Here's the thing. He just could not make it today, but he will be in it moving forward and he's in the Ord will continue to. He's also added some files to the repo. The sick clans repo, like about client libraries and some of the stuff that he's already working on. + +Greg Osuri: Oh perfect. So I'd love to see I was actually funny enough. + +Joseph Tary: Yeah. + +Greg Osuri: I was actually working on a golang client for Akash. A brand new client. I was calling it Eve and + +Greg Osuri: I mean, I, you know, I'm like, I'm known in the space for writing times like like Hashcore Nomad, for example, use Akash like CLI the code for for Akash, right? Hello communities, help. He uses our libraries to actually render the helm itself. So we have like pretty solid like, like, portfolio off, like, libraries that I want to incorporate back into like a customer online so far, the command line. We kind of kept. It low lifts to, to stay in sync with upstream because upstream gives us a command line library. So, I mean, the cosmos SDK, but now we had a decouple, the blockchain aspect of it and actually build a real command line, right? I think it's + +Greg Osuri: You know, for someone that wrote code that's still in use by Hashicor and Kubernetes and whatnot, not having a decent command line in Golang, it breaks me every day. + +Joseph Tary: Yeah. + +Greg Osuri: I cry. Okay, right before I sleep, but I want to change that. So ideally I want sort of like, I, I would propose reading working group for like, go to my line. Go Go library and some online along with the Goal library. So that becomes like the imports for the terraform libraries, that we had went a lot of interest from, like, hash account. People that wanted to, like, actually promote the telephone library, but I feel a little embarrassed because the telephone library is still depends on the command line. Not, it's like a rap online. + +00:45:00 + +Joseph Tary: It's not actual Lib. + +Greg Osuri: It's not, you know, actual command line, you know… + +Joseph Tary: Yeah. The other language I was thinking too would be good at binding. + +Greg Osuri: what I mean? Like, + +Joseph Tary: Forward would be a utility library for Python for people who are doing ML and want to start using. Especially once we have GPU support, one Strp able to use it for that. If we can, just give them like a simple thing, they can pull in through PIP and just be, you know, basically five or six lines to get a deployment with GPS for it up on a cash. I think that Bcpr useful for people, + +Greg Osuri: Absolutely 100% agree with you. I exercise like putting a lot of work on Python. I don't like a little book for 11,000 words here, but I'm a big fan of Python. I really also Python has sort of like, amazing user experience people. For some reason for command line experience,… + +Joseph Tary: Yeah. + +Greg Osuri: right? Like the, The Dot Opt group in Python, they're like they wrote the first like the, you know, like this is on how command line should be designed something big fan of so you I would like propose from priority standpoint golang, Python and JavaScript as + +Joseph Tary: Yeah, I think those are the core ones that people working with right now. + +Greg Osuri: We also have Dan Lynch, who's the founder of cosmology and he's you know the Britain quite a lot of like Cosmos Osmos and all of them use the their libraries. Now he wanted to, you know make some recommendations and propose some some you know. Items for our, for the, our akashi as itself. I think, is worth taking a look at the proposal as well, but we're starting to have like a lot of people coming, you know, forward wanting to like contribute Auto Akash depot. So, that's something that we need, we could consider when we are, like giving this JavaScript bindings, right? Cool. + +Joseph Tary: Yeah. + +Greg Osuri: Awesome. It's exciting. + +Jeet Raut: I just wanted one more thing. + +Tyler Wright: GG, has something. + +Jeet Raut: Yeah, thanks. so boss, we shot to me and we'd spoken a bit at a cosmokers about this potentially like the use case of ISD for Akash For settlement, if that's until something you guys are interested in. + +Greg Osuri: For settlement. IFC, absolutely. + +Jeet Raut: Yeah. + +Greg Osuri: Yeah. Oh yeah. Oh sorry. Yeah, it's good. That you mentioned. Yes, absolutely. So that's that's a big economics looking at We want to enable. + +Greg Osuri: A multi-currency settlement mechanism, really want to enable skate with mechanism, but we are not calling stable. We're calling a multi currency because technically, underneath the herd, you're actually opening up for any type of currency to be as a settlement layer beyond Akash. + +Jeet Raut: oh, + +Greg Osuri: Right? And the proposal is, well, you got to do that without undermining Akt, because Akt is integral to the security of Akash network itself. + +Greg Osuri: So in order to do that, we can incorporate IST, for example, but charge attacks, right? That's determined by the community. In the initial phase is tax non-vember that aggressive or like obviously going to be growth focused instead of value extraction focus but having a mechanism to extract value from multiple currency usage. I think is beneficial from the get go instead of getting surprises to people and so to encourage more people, we can be more growth oriented and have a tax rate to set to zero in the initial phases to encourage more people to deploy using as you're, right? So But ultimately the goal is to to create a both solve the user experience without undermining a kitty token. I think we have we found a way and I'll be proposing that in in SIG. + +Greg Osuri: Sick economics by End of the Week. Also submit a proposal, at least on my thoughts are. Rough proposal, wouldn't call like a panel one, but at least what my thoughts and how we can incorporate that absolutely love to mention your ISD in the paper itself. So we can put some gas on it. And another, another way we're looking at is so obviously it's gonna be a tax. That is maintained in the buffer as an IST. That can be further used to incentivize development. It's already Devon, right. Or incentivized robot proper providers to give them subsidies in some cases and we're also looking at a buy burn mechanism. + +Greg Osuri: So we would use some of those funds that from that we get from the tax to actually buy Akash directly from inside from a decentralized exchange and… + +Jeet Raut: You. + +Greg Osuri: burn some of that to a crew of value for NON-STAKERS, right? Um, so so that way, you're actually increasing your KD value by using it IST, and you can make it solid case for that. So, yeah, yeah. So that, yeah, super super excited. I think that's something we can. It be great like to collaborate on like,… + +00:50:00 + +Jeet Raut: If? + +Greg Osuri: enabling AST as well. Even open to I'm even open to enabling Asta the first currency before usdc, right? Like, because beneficial and ecosystem like synergies there, Because I'm a big fan of Algoric. So, + +Jeet Raut: Yeah. No, thank you or a big kinds of you guys as well. Yeah, I just let me, you know, getting bothers me on telegram and get you as well. Just let me know what you're thinking on that front and as well. Again, if there's anything And it's very interesting and thinking about me for this specific thing as well. Definitely interesting. Hearing other people's problems as well and like If there's anything on the smart contract side, that's something that we can work with, as well, with you guys, like if there's anything,… + +Greg Osuri: Yeah, maybe a lot. + +Jeet Raut: like Yeah. + +Jeet Raut: Yeah, right. So, + +Greg Osuri: Yes, very lot as a settlement comes closer. + +Greg Osuri: We're gonna have obviously, this, some work this world in terms of smart contract right now, I think we have to enable like cosmosom on a cash at this point to Or like are How's Agoric? + +Jeet Raut: Right. Makes sense. + +Greg Osuri: Smart contracts. Is there a module? That's portable. How does it work right now? + +Jeet Raut: let me, let me ask for a product because my understanding is right, like, we don't have Zoe's or Smart contracts. So Endojis we have a custom VM so it's I don't think it's quite as portable as everything that like Cosmos is built out. I'm assuming it I'm assuming so because otherwise up most, everyone would just use JavaScript is See, right. So + +Greg Osuri: Fair enough. + +Jeet Raut: But I'll ask you on that. + +Greg Osuri: I, Caused him is amazing. As much as amazing. It has not finished set of problems. + +Greg Osuri: Have you but we'll take it off, right? This meeting is public but I want to reach out to offline. + +Jeet Raut: Yes. + +Greg Osuri: I think we have some ideas on on improving but yeah, let's keep the scope for this meeting. Just sticky plans. + +Tyler Wright: Yeah, I think I just put in the calendar. + +Jeet Raut: Yeah. + +Tyler Wright: So again, that calendar should give a list of all the CIGS and all the various working groups. So again if you want to participate in the sick economics or any other, CIGS feel free to add those to your calendar, anybody that participated this meeting anybody that has access to sick and also invite other members to sing. So if there's anybody else that you think should be a part of this conversation, please feel free to fight, invite them to the monthly meeting and then again, we have a discord is another place to continue conversations between meetings. So I know that some folks like this here on folks are going to put some information and Max, I'll work with Max and make sure that some of the data that Greg you're talking about, could even be available before the next monthly, meaning the discord. And then, we'll obviously also be putting the Meeting Transcription Action Items meeting. + +Tyler Wright: Recording in the GitHub so that everybody can have access to the files. + +Greg Osuri: And also a little bit more publish on our website to Akash and Android slash community. I think we have like the SIG clients. + +Tyler Wright: Yep. + +Greg Osuri: Client 6, Fishing Interest Group. There you go. So, this should be the URL. That will be updated momentarily after this meeting. And moving forward, all the SIG meetings, all the transcripts, everything is going to public and you know, like available on the website. So easy, let's index by my Google and You know it's also available machine format, so people want to sort of like incorporate the meetings in their own lighting. you know, I wonder but if you want to + +Tyler Wright: Cool, does anyone? + +Greg Osuri: And I'm proud of new website to be like super like up to date with everything that's happened in the community. + +Tyler Wright: Absolutely. Does anyone have any other questions or comments that they want to share today? And this initial SIG clients meeting before the chocolate? + +Tyler Wright: Cool. I would also add that because oh go ahead. Go ahead. + +Jeet Raut: Other quickly. Yeah, I know. Thanks, thanks so much for having us. But yeah, Greg off all these separately. + +Greg Osuri: Sounds good. Thanks you. + +Tyler Wright: Cool. I was just going to say if anybody else wants to We're involved in any specific SIG and some sort of leadership role. Please feel free to mention that discord because this is open, you know, obviously we're chartering this thing but eventually anybody can kind of lead the structure and kind of coordinate. So please let us know how you want to get involved, but we really appreciate the time today. + +Greg Osuri: Sounds good. Thanks everyone. + +Joseph Tary: And thanks. + +Tyler Wright: All right. Thank you. + +Andrew Gnatyuk: Thanks. Bye guys. + +Tyler Wright: I could. + +Joseph Tary: But everyone. + +Meeting ended after 00:54:51 👋