Fluence Labs Community, 24 Feb 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Developer Community Call #10 – Fluence Registry

Description

00:00 Intro
00:42 What's new in Fluence. Latest updates
04:01 Hackathon Bounties
05:16 Codeless Conduct Hackathon& Bounties
06:01 We are millions Hackathon
06:58 Tech updates from Dmitry Kurinskiy
13:00 Presentation. Fluence registry
13:11 The content of the presentation
13:44 What Fluence looks like
15:24 Fluence Registry in a nutshell
17:26 Fluence Registry: a design approach
17:38 Routing in a decentralized network
18:40 Kademlia algorithm
21:00 Kademlia is not a complete solution for the routing problem
23:20 Lifecycle of routing
24:58 Routing data. What data is needed
26:55 Distributed routing. What data do we need
28:47 Routeld-demystified. How to register it / to it?
31:16 Registry Route. Data types.
32:49 Registry Security
34:55 Possible use cases
35:35 Q&A
37:57 Demo
42:31 Q&A

Speaker: Dmitry Kurinskiy

Useful Links
https://fluence.network/
https://twitter.com/fluence_project
https://doc.fluence.dev/docs/
https://doc.fluence.dev/aqua-book/
https://github.com/fluencelabs/aqua
https://github.com/fluencelabs/aquavm
https://fluencenetwork.notion.site/fluencenetwork/Fluence-Developer-Community-Calls-2733beddb3ff4e4986c4fcc830fc796d
https://github.com/fluencelabs/marine-rs-sdk
https://t.me/fluencedev
https://fluence.chat

A

All right welcome everybody february 24th, another edition of fluence community call and we got a pretty packed agenda today. There's some great news about uh fluents. I think if kenny has to share, uh we have dmitry with uh some updates and uh some pretty exciting uh demo and uh we have uh hacker who's uh representing his project so to kick it off afghani, maybe what's new influence.

B

Yeah, hello, hey everyone. um I guess the the biggest update since um last call is that we officially um announced our fundraise uh nine million dollars round, led by multicoin capital, with participation of others, great funds and individuals. So we are really proud uh announcing it really proud having this funding secured and yeah. So this this is exciting. uh We are uh published like basically, the goal of this fundraiser is to accelerate the uh building an ecosystem development.

B

So we are published um a lot of open positions, engineering roles, product marketing, uh developer, evangelism, roles, engineer management, so um um really trying to scale the team right now um and we are with this round. We are getting closer to launching fluency down, so this will be the next big step for um our community.

B

um So currently it's just fluence labs who develop in software, we'll be creating dao and basically the dao will be managing the the network.

B

Influence labs will be participating um in in this, alongside with the other community members, investors and partners, and everyone else so um working on um right now on like on preparation to the dell launch, um but on the development side also like we continue uh doing hackathons bounties, um and you know happy too happy to be at this stage right now um um and the other important thing is: uh um we've been in denver last week.

B

So denver was pretty cool, uh a lot of people, fifty thousand fifteen thousand uh people there um and the hackathon well, most of them actually was not for the hackathon. Most of them were just like just meeting each other, so it was pretty um pretty nice.

B

um Great time met a lot of people um and but the the offline in the hackathon, I think, already ended on sunday, but the online in denver part is still alive and I don't remember which exact deadline, maybe bernard, can come up with his dates, uh but um so there are still bounties uh published by us for this hackathon they're, still life and still opportunity to participate in that um yeah. I I think that's it. uh That someday is on my side.

A

Yeah with respect to the the virtual side of the hackathon, it goes until I think the end of march march 24th, or something like that. If you look at the repo github forward, slash fluence you'll see some slaps. You see the uh uh repo for east denver with all the bounties. I think we have three bounties. There is uh a lot I think up to fifteen thousand dollars still available, and um it's actually some pretty cool pushing. I don't have in front of you.

A

I have to look them up, but they're pretty good and uh if any one of you is interested hit us up in discord telegram whatever we'll happy to uh help you guys with the rest of the virtual hackathon, which we expect to be actually uh pretty sizable.

A

Having said that, the in person I think, had over 100 teams, 100 200 teams- and I don't know how many submissions it was uh crazy. So we expect nothing less for the virtual one uh dmitry, I think, has some updates on the tech side.

B

Should we just quickly also mention this uh coldlights hackathon, which is yeah.

A

Okay, so I was okay, I was gonna have me to do the update and then anna do an update but uh sure uh anna. Why don't you tell us about the upcoming uh hackathons.

C

Sure um so one hackathon, which the name is codeless conduct, is the hackathon for those who um who are non-techies. um Those of you guys who are super creative and good in design or writing articles or making videos or nfts.

C

So we have, our bounties are like the pool. The prize pool is about 10 10 and a half thousand, um and several bounties have multiple winners. So this hackathon has already started.

C

You can apply through git coin and participate not only in our bounties. There are also um a common track and a lot of interesting stuff.

C

um That's for codeless hackathon and then one other hackathon, where we will be starting it's on first of march, it's the hackathon called. We are millions together with the swarm, so the bounties will be announced soon and we will.

C

We will also notify everyone in discord and that are the two two hackathons that we will be running besides denver one.

A

Okay, so lots of hacking going on both the text and non-text uh if you're interested or anybody. You know interested all the all the bounty listings and uh judging and submission instructions are going to be on github easy to find forward slash, release labs forward, slash from the slabs you'll find it there all right, meetry.

D

Onto you so uh hi guys, I have some some updates and uh I have a talk, and ideally there should be something in between uh the updates and and the talk.

D

So, uh first of all, we have a lot of things done uh last last two weeks and uh delivered today or yesterday, like just recently uh and uh the best way to learn what exactly was delivered is to subscribe to our updates channel on telegram.

D

Hopefully we will be cross-posting it to discord as well, but currently everything is on the telegram updates channel. But uh what I really want to mention is uh the first milestone of marketing web.

D

uh It's our efforts to run webassembly interface types compatible web assembly services in javascript environment, so the first milestone is something in between the proof of concept that interface types uh really work inside the web browser and in the javascript environment.

D

uh But it's like slightly more uh because it's actually currently used in production in the in the fluence javascript implementation.

D

So currently you cannot use marine web yet uh and the only module that actually works with marine web is a arc by virtual machine. But if you download and use uh fluence.js 0.20 and any later version, then marine web is under the hood in the following weeks.

D

uh We're going to reach the second milestone, which is simply providing a public api for the developers to run web assembly modules uh with interface types in in the browser and on the javascript, fluence beer implementation and the third milestone will be supporting multi-module services, so we will implement linking for marine web services for javascript, just like just as as linking works for web assembly services on the rust node implementation.

D

What does it mean for the developers? It means that if you want to use some complex logic you you can like, you can use it. Despite the environment, you can choose to use javascript or rust or some other environment that is compatible with the node.js or with rust, and you can just use it. If you develop a module, then it can be used in any environment.

D

It makes the network more isomorphic and in the following weeks and months it will be heavy used by the fluence protocol. So this is a very big update.

D

That that we have, we also have some minor updates with aqua cli to simplify the life of developers. Now you can deploy schedule scripts with a aqua cli, and this functionality is much more rich and useful than it was in the previous flist.

D

So take a look on that and finally, we deployed delivered released the fluency registry that I will have. I talked about soon, so I'm done with the updates overview. If anyone wants to ask anything or add anything, maybe my equilibrium alexi, maybe anyone.

A

No questions from anybody.

A

Okay, why crowd today.

A

I I okay, I'm not just informing that we, the hacker or the team, the participant of previous pakistan who was gonna present today, is not presenting so dmitry. We don't have a buffer for you.

A

Let's go straight into your presentation.

D

Okay, uh so I will, uh I will share.

D

The presentation.

D

Can you see it.

D

Yeah yeah sure.

D

Fine tuning okay, so today I want to.

B

D

The fluency registry, the default routing level uh of the fluence network, so what's going to happen uh first, I will show what fluence register is just in two slides with the code snippet just to show what it looks like.

D

If, if you want to have a fast line, then I will discuss the design and the reason why registry designed it this way, but it's implemented this way and not another way and finally, hopefully we'll have a demo.

D

So first I want just to remind what fluence looks like we have this picture: a wonderful picture that we have some aquascript, that flows from one client from the web, browser which is not publicly accessible, but it is a fluent sphere to some node to another node.

C

To another node and.

D

On some node it could run different services and get get data back from these services, and then it could reach some other peer uh like another browser or it can pin back uh to the caller. So it's pretty nice, it all happens without central coordination.

D

You can describe protocols as aqua scripts and then compose them compose services, apis protocols, applications using arkwa and everything should magically work, but when it comes to reality, the first question uh that is raised is how exactly to choose where these arrows should go like how to understand to what node uh the script should flow, uh how to be sure.

B

D

For example, the client, uh despite the the relay node, is connected to is reachable uh and can get a notification. So how exactly to make this topology is the problem of routing and fluence registry is a solution. uh The default solution not on not only not only possible but the default, and uh this is what it looks like uh to use registry. You need to have something on the host language, for example, on the javascript side, and you need to have something in aqua.

D

As a registry is a kind of protocol and to compose protocols, you need to use aqua.

D

So on the left, we have a creation of the flynn spear to to have the the key player to have a public and private key to secure the route. Then you can create the route and to have the root id.

D

And finally, if anyone register registers itself to this route, you can send a message to everyone who subscribed- and here we have some aquacode. So we have that data type, a custom data type for event.

D

We have some service, that's going to be launched on uh uh on the remote computer and we want to deliver this event to all the registered peers of some root id.

D

So we resolve the root, uh it's library function and then, for all the records for this route for every peer registrat, we can have a relay. We can have a peer id. uh We can understand what exact service we want to call and finally, we can call the service and hence deliver the message.

D

So this is what the developer experience looks like and now I want to go deeper into the details and the motivation why why is it designed this way? So it's about the design approach.

D

So uh what does mean to have a routing routing in a decentralized network?

D

So essentially we want to run some computations on a particular peer, given that the network consists of peers, including not publicly accessible, peers, client peers, and these peers are different- they have different resources, they have some deployed services which you expect them to run, or they could have an access to some data or to some capabilities uh or they could participate or not participate in your application, so they are different and on runtime, during the request, we need to find a peer uh to access his resources.

D

So that's the the problem that we want to solve and we are going to solve this uh uh problem in the context uh of existing algorithms um and the main one is kadamlia. Kadamla is backbone for many decentralized networks, including uh the fluence network, uh and uh it's super nice super efficient. uh It provides logarithmic complexity on the size of the network uh to make almost all the requests, uh which is pretty nice but fundamental name uh in the level of types you can think of kadamlia as just two functions.

D

First function is a lookup for any peer id. You can get an address.

D

If this peer is online known and rich blue, so for any peer in the network, you can do this, but if this peer is not reachable, if it's a client peer or if it's offline, then you will not get the address to connect to that's why the question mark the second function uh is called the kadamla neighborhood.

D

So for any hash, any nonce, any byte array you can find a set of peer, ids or a set of peers called the neighborhood of this hash, and this neighborhood is consistent, but not not not predefined. It's not consistent in the meaning that you know what exactly you get for. This hash is consistent, meaning that probably for every period on the network, this neighborhood at the same time, will be similar or the same.

D

So that's why you can use this neighborhood for eventually consistent communication, but you cannot decide what peer will go to what to which neighborhood.

D

And uh that's why uh kadamle.

C

D

uh Is not a solution he's not a complete solution for uh the routing problem? um What what do you want to have?

D

Let's imagine that we have some root id and uh essentially what we want to have is a function that gets from this root id to a to the registered set of peer ids to the predefined set, so that we can get the same predis for the root id. If they are available, then we want to have uh connectivity to the peer. So even if the target peer is not publicly accessible, if there is no public api, if it's a web browser.

B

D

If it's the node with public api, if we want to be able to connect it uh in both cases, somehow uh the third problem is that we need to solve the last mile problem, which means that we need to have some routing logic inside the peer.

D

We need to get the service id that will serve our request inside inside the beer. So this is also what we want to have and uh also there's question about security. uh How can we protect this root id from from unauthorized rights? uh Should we do it at all? uh Should we do it in all cases? How exactly we want to protect it? uh What security level is enough or not?

D

So it's a question and finally, I'm speaking about root id, but the question is how to get this root id and what exact meaning it has so that's kind of an application, specific logic, but it needs to be solved somehow to find out how to solve it. Let's take a look on the life cycle of royalty. uh What happens on the way?

D

So, first of all, we have some criteria, application, specific criteria, probably, and we need to decide what route to find we need to have this root id uh to look it up later, then, we have a distributed.

C

D

We need to resolve this route to get target peers and we need to deliver a particle to the target, pure in the terms of aqua it needs. It means that we want to state that we want to execute one target pair pure, and we expect that we want to get there via direct relay and then it means that if target relay is publicly accessible, then we can execute the code on the target period.

D

And finally, we have the problem of local routing that happens inside the pure and in terms of aqua. It means that we need to resolve some service that we care about, provide the the local service id for this peer and, finally, a call function.

D

We assume that you know what function you want to call and you express it in types. So, the last the last step is solved by the type system.

D

uh During this life cycle, we can understand what exact data is needed to provide routing capabilities.

D

First of all, it's clear that we need to resolve a service and how can it be done?

D

How can we resolve service?

D

We might take some services id and assume that it's given it's a part of the data provided by the rooting algorithm, but for some algorithms or protocols it might be not not the case because you might have a predefined, alias the name of the service, which is a very frequent use case for the javascript code in the browser. Usually, you just register some service with a known id and it's pretty fine, you don't need service id or you can run some local selection processes uh to find the service inside the node.

D

uh When you got to the node, um it could be any, but it's also the key is so it would be nice to have an optional service id and for the remote part for non-publicly, accessible peers. We need to have relay id a publicly accessible peer that we want to deliver the message before it's delivered to the target purity and for sure we need to have peer id.

D

So that's two fields that we that we need.

D

And that's how it could be expressed in some kind of a certain code. We finally want to have the distributed routing in form of something like this.

D

You want to resolve root id into a collection of data structures with an optional relative id or service id, and actually it's super useful to have an optional value to filter, appears or to have a vector clock in this value or to have some some nodes.

D

So it's also super heavy, and if we have such function uh then we can have academia. We can have this function deployed locally. We can have a service with this function and then we can use a lookup from condemned to oh, it should be inversed.

D

uh We can use uh neighborhood to decide on what peers to repre distribute these data and we can use the lookup to resolve, addresses of relay id or peer id uh and to deliver the message uh to have connectivity and finally, it will be proportional to logarithm of the size of the network. If not cached.

D

We can provide some cache for this request as well.

D

uh So essentially, this is what we have in the in the service. Slightly more complex, but this is the meaning.

D

Finally, we need to think about the road id and demystify it how to register it, how to register to it.

D

As we have an open promotionalist network, we need the root id to be a random hash, to avoid collisions and to ensure immutability of the roots uh metadata, and the question is what exact metadata we want to have and what? What exactly we want to protect from changes. Registry provides uh one default solution: it's not the only possible, but with registry, every peer can have an isolated namespace in this form. So the root id is the hash that depends on the root owner id and the label.

D

A label is an arbitrary string, so uh you can construct construct, build the road id if you know target's purity uh and the value of the label. For example, you can say that the label for your experience-based protocol is, I don't know, um user data and then, if you know peer id, you can try to resolve these. Your ids use your data or user profile, for example uh by hassan this. uh This fields, his peer id and the string user data to get the root id to derive it from from the data.

D

You know, and uh one important outcome for all this system is that there is no way to pick root id of your choice. You cannot create the root id like. I don't know uh that that's called my app because it will be hashed just like you cannot upload different files to the same cid in ipfs or have two different names on ips.

D

That's just the same mechanics.

D

So. Finally, uh this is our the the data types that we actually have in the registry implementation. There are two structures. The first one is a is root.

D

uh Here we have the root id uh that is returned by the service or by the protocol to you uh is derived from the label and the peer id.

D

uh Then we store timestamp created why the restore times the time created and why it's not the part of the root id, because you can it it's a very, very simple way to be able to update the metadata for the root and the metadata is challenged.

D

And finally, you have a signature that proves that all these fields were set by the spirid, and then we have the record that references to root id that could have some small value. uh Its points on the peer id is set by another pid.

D

You have relayed these service id timestamp, created solution and the signature signature secures all the fields above.

D

And the last question is what about the registry security, so first of all, both road and record are signed, so you cannot mutate any other peers record or you cannot register it for another user's behalf.

D

So it's part of security and the discussion mechanism is also part of it and then, as you probably noticed, the root data structure contains an optional challenge, challenge and challenge type and every record to be added to be registered for this route must contain a solution, so you can protect.

D

You can have some security checking for for authorization for rights into your your route for registrations, uh and we have the default implementation for changes and solutions is called trust graph and actually it's available. Now there is no announcement yet, but it's published and probably will have another dedicated community called to discuss trust graph and, as we have a lot of working with the signatures here, uh there is a sig service uh which is used to manage client signatures and key pairs, and it is also available. uh I'm not sure if we have the announcement.

D

Probably we should have it uh today or tomorrow, but the code is the life. It's used so we'll cover it also in in a different call.

D

I forgot to write down the use cases, but you can imagine the gossips the chat, rooms, mutual discovery, discoverable user data or many devices for single peer without sharing the key players has on and now, let's switch to the demo time.

D

If there are any questions about the the theory is good time to to ask, if not, let's switch to the demo.

D

Sorry, can you repeat.

E

D

Is delivered with the fluence js and it's available.

D

We just need to merge documentation, uh but actually you can learn how to use it from uh from the demo code that I will present or from the graph code that is available, but we will describe and have a dedicated announcement for six service trust graph is uh available. There is a javascript demo for trust graph, but we haven't shown it because these are two sides of one metal, uh truss graph and registry and.

C

We decided to show.

D

Registry first, because this roading problem is more useful, while trust graph is actually a very rich concept and very fundamental for the network security in general, but without registry, uh it's just some signatures uh and there's nothing to show.

D

That's why I decided we decided to postpone uh the demo of a trust graph and to show registry first registry was deployed on crosstalk network today. So it's live right now it might be slightly not stable.

D

Maybe we will be fixing some some minor bugs in the following days, uh but it's live.

E

Yeah, okay: I thought it was available, so it was live. Yeah! Okay, thank you.

D

So, let's, let's switch to the drama, I hope it will work.

D

So what I want to present.

D

I have created a fluence here. I want to write down my name. uh My name is my name and now I will register the the root.

D

And here, if you opened this link with your phone or anything like that that, if, if you scan this qr code, then probably you'll discover that I exist. Oh and here we have someone else and uh if anyone else joins here, uh he will discover both of us and will appear here yeah. So hopefully it works. uh It creates it created three roads right now for every device.

D

There is a registered record for the relay and for this device, and uh they can learn about each other by asking me and they find the way to me with the.

D

The root id- that's basically it we have more and more people here, that's nice!

D

You can take a look on the code uh in uh the registered demo repository in the fluence labs uh and that's all I have thank you.

A

All right, thank you, dimitri questions. Anybody. This is super exciting.

D

C

D

That I want to mention uh we have this notation for uh discovery of the root and, uh if you think about it, it's super powerful uh and uh this um this layer can be used for many purposes, including quite complex scenarios like gossip or data synchronization, managing replica assets, and so on so hope uh we will be showing more different demos, which will finally have roots just in these.

D

In these data types, which are super super rich.

D

Any questions any reaction: if there's none, let's, let's go.

D

Woohoo, here's the link to the demo uh in the chat.

D

So if, if you want, you can take gluten.

D

A

Thank you, okay. uh Nobody has any questions. I think that pretty much concludes does anybody else want to add anything anything you've been working on. You want to share, or particularly excited.

A

Really, that's no share environment here, it's anything but collaborative uh okay. Well, in that case thanks everybody, so me too, when when are you expecting to demo the trust graph in two weeks.

D

Well um it it should be absolutely possible, so we aim at uh hang the demo into weeks. Okay and.

A

D

Probably we will have it connected to the red string at this moment.

A

All right so, uh and anybody who wants to use it uh reach out and telegram discord any channeling. You want and we'll be happy to support it, and I expect it to start popping up in uh upcoming hack of sounds as well, so get a head start on it. For those of you on a call. Definitely it's going to help you all right. Well, see you in two weeks thanks everybody!

A

Thank you. Bye.

D

A