IPFS Weekly Call, 6 May 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IPFS Weekly Call 2019-05-06 🙌🏽📞

Description

IPFS Newsletter: https://tinyletter.com/ipfsnewsletter

A

So you have the crowd here.

A

You, okay, awesome.

A

Now, let's start recording.

A

We're live, yeah are certainly long. Can.

B

I do a quick plug for the office. I was cool them. Seeing us now we're on video, we're.

A

Live and now we're recording so hello, everyone and welcome to the ipfs weekly call, but we get the scene of great things that are being built on top of IP FS. Today we have a presentation about keep Network, but before we begin before main presentation, Ali has an exciting announcement. Apologies.

B

For the people on the call you know that is three times but ipfs office hours is a thing that happens just before the idea. That's weekly cool, so you missed it because this weekly pulls happening, but it will happen again next week and if you're interested in contributing so I give this, but don't know where to get started or you're running in a production, and you got some questions like we're there to help its. We are available to talk to you come check anyway. Thank you very much. Awesome.

A

Once again, Ali thank you for doing this. It's great all right, so we don't have any more announcements, so we will start, and today's presentation is about network tape. Network is an off chain container for our private data and we have block ha if your chain here, to tell us and talk about keeping Network and how its using lid p2p. So, if I further ado ah václav appeal to thank you for being here and I will turn it over to you, so you can present Thank you Thank.

C

You Portia thank you for inviting us. You're really excited to be here review. Can you see my screen? Yes,.

D

C

Awesome so my name is Pioche together with raga. We would like to give you a presentation about how do we use the p2p keep network? We will first talk about what is key and how it works, and then we talk about how do we you slip it to pee so jumping to keep if I had to describe keep in just one sentence, I would said that keep the privacy liar for public dog chains over the last few years, public blockchains pros incredible transparency and auditability with in the table and censorship resistant records.

C

However, any smart contract that is published to the blockchain can be easily accessed by competing interests, so when companies consider building applications on blockchain privacy is one of the primary issues that arise and keep is a bridge between public blockchain and private data, so public contracts can held as the full power of blockchain technology without compromising on transparency and privacy. We can also say that keep is a form of a private computer able to store and compute data hidden even from itself.

C

One of the underlying primitives of keep network is the random beacon which is responsible for providing a trusted randomness source for public blog chains. We can divide, keep Network into two parts. The part that we call on chain is a set of aetherium contracts that sell as a kind of an api to our network. The second part that we call off chain is a connected network of keep clients which use special communication protocols to serve requests from the chain.

C

For now, we can focus on random, become because, from the perspective of the PE, we will use pretty similar mechanisms for all other futures of hip network. So the way when the beacon works is that external client publishes a request to the chain. The request says: please produce a random number for me. Keep clients receive this request from the chain and subset of them. Those clients that are in the group, which is the chosen group, cooperate to produce a random number.

C

The way this cooperation protocol is designed does not allow anyone to employ the result, and the result can be later. Cryptographically proved to be free of manipulation, so keep coins published the result to the chain and they start another protocol called group selection that will allow them to elect a new group, and this new group will be added to group registry and this group, this new group may or may not be asked in future to produce a new random number. This is also kinda random, which group is chosen.

C

So if you look at it from off chain perspective, this part that is connected to the chain. We have a clients connected to each other using clip p2p inside this network, clients form groups. One note one pill can be a member of several groups: clients, peers, communicate inside those groups. Groups are chosen randomly and they have a limited time span from the perspective of the network. We need the following properties, so clients must be able to efficiently exchange messages inside those groups. Messages must have a very, very high delivery rate inside the group.

C

Some communication, the group has to be public. Some communication has to be kept private between two clients. We require message attribute ability, so sender must be well defined and the message must provide senders of the station. We require integrity of messages, so we need to be able to detect a message was tampered and we need a pill.

C

Discovery mechanism that, when a new client joins, join, join the network, so lead p2p and specifically the purpose of implementation and build up being cover that ended up being this decided as a best option for us, mainly because client is written and go our clients in training go and the p2p was actively developing, go ipfs and Goldie p2p suit off libraries.

C

No rocket will present how exactly it looks under the hood.

E

Right, like so Peter said when we started this adventure about like a year and a half ago, the options were really led: p2p go with areum's discovery code that they have and then rolling your own, and we decided to you know, join, would p2p and ipfs because they had quite a bit of traction and so how we use it in our code is that clients come in with an aetherium account.

E

These are ECDSA keys, and so we have interfaces between our aetherium code and our live p2p code, and we take these ECDSA keys and we turn them into the Lib PDP representation and then from there we get our peer ID and we also use listen address. We use DHT hosts custom transports, so we instantiate all this stuff and that kicks off our network that allows the client to connect to the lid p2p network and then going to the next slide.

E

You'll see that what we do from here is that clients need to find a bootstrap list, and so like well table show a bunch of bootstrap nodes, we're a little worried about this, because they seem like boss, vectors but we'll publish these nodes. Clients will connect to the bootstrap nodes. One of the things that we need to do is ensure that there's a good diversity of bootstrap nodes that clients can connect to also retries, and things like that.

E

One of the interesting things that we got it to do is that we wrote our own custom transport to hijack a connection before the key client joins a network that would p2p network. What we do here is we get the connection.

E

We check the PA look to the hear each other stay chained. Then they do a bit of a handshake like a little bit of a challenge response and once they've authenticated each other, like okay, you're good to go you're good to join the network.

E

So Peter mentioned that we have a group selection protocol in that group selection protocol clients basically submit tickets on the chain, they're validated like okay, these tickets are good. This is how many shares you get and that allows you to join the pub/sub channel and so basically the pump sub channels. Just like a you know, a randomized words just a hash of all the clients and they join this channel and every group is its own pubs own channel. This is a topic and go with PDP pub/sub.

E

So we just like form that topic and we currently use flood sub. We've played around with gossip stuff, but we want to initially go with flood song so that there's a good. You know diversity of messages hitting all the peers, and if we see that episode of gossip sub are better for a use case in the future, will do that.

E

So sending a message- and these the next few slides are a little bit out of date because we updated the p2p to take advantage of the strict signing and verification. So what we do is we just you know we get this protocol message. Whatever protocol, the clients are executing, so there could be dkg or relay signing and what we do is we marshal be specific protocol message. We like register sub handlers and then we just send it.

E

We just call publish we no longer have our internal signing code, we just got it all, let out and we're relying on the PDP for signing the message and verifying it, and so this is what the message layer looks like again. We use protobuf you'll, see the lid PDP pub/sub message, and then you can just ignore the network envelope, because we removed all that and now it's as simple as the network message itself.

E

Right and so for receiving a message, we basically call next, which is a method that the pub/sub interface exposes and every message is basically treated like an HTTP request. We're like okay, this is a mute, go routine process that message and then call the correct handler to send it to the correct protocol that executes.

E

Right, and so we also wrote a little bit of custom code, for example, handling the different key types, but again we primarily focus on ECDSA and ensure that they're interoperable with each other, and we have some custom handling code as well right, so things that we need to do or that we could use. So one of the things is that we need deep connection management. I believe this is going to be possible with the connection manager, but I'm not entirely sure.

E

The reason we need this is to disconnect clients that fall below the stinking threshold also, if we need to disqualify view there's a given round of the protocol. So we need something that can let us like Band connections or cut them, or you know server them for some period of time. We need more resilient bootstrap, as mentioned before, so that's to ensure that we get a good diversity of booster appears and then we need to have a retry er there's a few options here.

E

We could have something as simple as just like retrying code, or something a little bit more complicated like what Hashi corp implements and swim, and the last thing is this: rework validation, pipeline and Golub PB pub/sub, so currently we're anchored to a commit that is in PR review.

E

This is because, after updating to strict to enabling signing within Golub p2p pub/sub, we noticed that some of us could run our end end code, our experiments and some of us couldn't- and we noticed that, once we enabled signing messages had a higher frequency of dropping for some folks on our team, and we finally realized that it's because validation what's happening in line- and this commit right here- splits it up into synchronous and asynchronous validation, which allows messages to be processed.

E

So expensive things get pushed off, async and then basic structural checks happen in line, and it would be great to get that version ASAP so that we can be back on master yeah. So thanks, that's our implementation. Hopefully that was there's enough information.

A

Awesome. Thank you very much.

A

A

Okay, um next is QA, so if anyone has a question, they could put it in chat. That would be wonderful.

A

So there are any questions regarding um network and Lupita Pete implementation.

A

um I have one question: could you tell us in like a higher level, who is using like a peak network like whenever building a smart contract and they need to network and they need to have a public contract, but a private container of data like water to use cases that you're, seeing initially.

E

So we're going to be launching our private test net pretty soon here, so we haven't launched yet, but the folks that were looking to work with our folks lot, who need who need these private containers to, like you know, Aragon or a few folks like that, are pretty interested in these use cases. Folks that need decentralized, custodial solutions would be very interested in this. This technology, one of the keeps that were looking at in pewter, is given a few talks on this that you guys can find online. Our Tec DSA keeps threshold.

E

At least DSA keeps that's going to be our first type and we're actually building our first product in that which will, you know we'll announce more on that later, so anybody who needs a decentralized custodial solution and anybody who needs private containers for data that could be interesting to mask some information from some period of time, but initially we're just focused on tec, tsa cubes. So anybody that has a private ECDSA keep that needs to be remote-controlled or kept securely on it. There him and.

C

Anyone who needs randomness in public blockchain, which is actually right now not very possible, keep Network make, makes it available.

A

Excellent, thank you. Do we have any other questions? Okay,.

A

All right, um if we don't have any other questions, then I. Thank you very much for presenting. We can leave it here. We can end here on everyone. You get about 13 minutes back and I will see everyone else at the next I can't that's weekly meeting. So once again, thank you very much for representing expression.

E

Thank you. Thank you.

A